US5859907A - Echo canceler and echo path estimating method - Google Patents

Echo canceler and echo path estimating method Download PDF

Info

Publication number
US5859907A
US5859907A US08/507,507 US50750796A US5859907A US 5859907 A US5859907 A US 5859907A US 50750796 A US50750796 A US 50750796A US 5859907 A US5859907 A US 5859907A
Authority
US
United States
Prior art keywords
echo
wire side
signal
voice
transmission
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
US08/507,507
Inventor
Toshiro Kawahara
Toshio Miki
Hirohito Suda
Tomoyuki Ohya
Hirofumi Takagi
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
NTT Docomo Inc
Original Assignee
NTT Mobile Communications Networks Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from JP11764094A external-priority patent/JPH07303067A/en
Priority claimed from JP11759294A external-priority patent/JPH07303064A/en
Priority claimed from JP11758994A external-priority patent/JPH07303061A/en
Priority claimed from JP6117590A external-priority patent/JPH07303062A/en
Priority claimed from JP11759394A external-priority patent/JP3212796B2/en
Application filed by NTT Mobile Communications Networks Inc filed Critical NTT Mobile Communications Networks Inc
Assigned to NTT MOBILE COMMUNICATIONS NETWORK reassignment NTT MOBILE COMMUNICATIONS NETWORK ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: KAWAHARA, TOSHIRO, MIKI, TOSHIO, OHYA, TOMOYUKI, SUDA, HIROHITO, TAKAGI, HIROFUMI
Application granted granted Critical
Publication of US5859907A publication Critical patent/US5859907A/en
Anticipated expiration legal-status Critical
Expired - Fee Related legal-status Critical Current

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04BTRANSMISSION
    • H04B1/00Details of transmission systems, not covered by a single one of groups H04B3/00 - H04B13/00; Details of transmission systems not characterised by the medium used for transmission
    • H04B1/06Receivers
    • H04B1/10Means associated with receiver for limiting or suppressing noise or interference
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04BTRANSMISSION
    • H04B3/00Line transmission systems
    • H04B3/02Details
    • H04B3/20Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other
    • H04B3/23Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other using a replica of transmitted signal in the time domain, e.g. echo cancellers
    • H04B3/238Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other using a replica of transmitted signal in the time domain, e.g. echo cancellers using initial training sequence
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04BTRANSMISSION
    • H04B3/00Line transmission systems
    • H04B3/02Details
    • H04B3/20Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other
    • H04B3/23Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other using a replica of transmitted signal in the time domain, e.g. echo cancellers

Definitions

  • This invention relates to an echo canceler suited to be used for a mobile communications network and a long-distance telephone line network. Also, the invention relates to an echo path estimating method suited to be used for this echo canceler.
  • the subscriber's line In a long-distance telephone line via a submarine cable or via a communication satellite, the subscriber's line, in general, connected to both ends of the line is of a two-wire circuit and its long-distance transmission portion is of a four-wire circuit for the purposes of amplification of a signal, etc.
  • the subscriber's line of a terrestrial analog telephone is of a two-wire circuit and its portion from a terminal of the mobile telephone to a switch, etc. is of a four-wire circuit.
  • the connection region between the two-wire and the four-wire is provided with a hybrid circuit for performing a four-wire/two-wire conversion.
  • This hybrid circuit is designed to match with the impedance of the two-wire circuit.
  • a received signal reaching an input side of the four-wire of the hybrid circuit tends to leak toward an output side of the four-wire, thereby generating a so-called echo. Since such an echo is lower in level than the talker's voice and reaches the talker after a delay of a predetermined time period, a speech hindrance is created. Such a speech hindrance caused by echo becomes more significant as the signal propagation time becomes longer.
  • FIG. 2 shows one example of the waveform of an echo with respect to impulse response.
  • FIG. 1 shows a schematic construction of an echo canceler which can be used in a mobile communications network.
  • the echo canceler 1 illustrated here is located on a front stage of a hybrid circuit 2.
  • the subscriber of an ordinary analog telephone is referred to as the "near-end talker" and the subscriber of a mobile telephone as the "far-end talker".
  • a far-end voice signal input into the echo canceler 1 is represented by Rin; a far-end voice signal output from the echo canceler 1, by Rout; a near-end voice signal input into the echo canceler 1, by Sin; and a near-end voice signal output from the echo canceler 1, by Sout; respectively.
  • the echo canceler 1 shown in FIG. 1 comprises an echo path estimation/echo replica generation circuit 3, a control unit 4, an adder 5, and a non-linear processor 6.
  • the echo path estimation/echo replica generation circuit 3 detects a response characteristic of the hybrid circuit 2 based on both the far-end voice input Rin and near-end voice input Sin and estimates an echo path (namely, echo propagating line). Then, an anticipated echo (namely, echo replica) from the hybrid circuit 2 is generated through a convolutional operation based on the result of estimation and the far-end voice input Rin.
  • This echo replica is generated by an FIR filter which is constituted of so many taps as 512, for example.
  • a convolutional operation in an echo replica refers to this.
  • this echo replica is subtracted from the near-end voice input Sin, thereby canceling the echo.
  • a learning identification algorithm is used. Among many adaptive algorithms, this learning identification algorithm is comparatively small in computational complexity and good in convergence characteristic.
  • the echo path estimation/echo replica generation circuit 3 has an FIR filter.
  • An echo replica signal Y(z) output from the FIR filter can be obtained by the following equation (1). ##EQU1##
  • N is the number of taps of the FIR filer
  • an adaptive filter technique for example, a learning identification algorithm, which among many other adaptive algorithms, is comparatively small in computational complexity and good in convergence characteristic, is used. Details of the learning identification algorithm is disclosed, for example, in Institute of Electronics and Communication Engineers of Japan (IECE) Journal '77/11 Vol. J60-A NO.11, article under the heading of "Regarding Echo Canceling Characteristic of Echo Canceler Using Learning Identification Algorithm".
  • a far-end voice output Rout of the level sufficient for an echo to come back as a near-end voice input Sin is present.
  • the far-end taker is currently engaged in speech.
  • the near-end voice input Sin is constituted of an echo (or an echo and a white noise) alone. In other words, the near-end taker is not engaged in speech.
  • the non-linear processor 6 In order to remove such an echo component (hereinafter referred to as the "residual echo"), the non-linear processor 6 is employed.
  • This non-linear processor 6 undertakes a non-linear switching operation. Specifically, in the case where the near-end voice output Sout is constituted merely of an echo, in other words, in the case where only the far-end talker is currently engaged in speech (this state is hereinafter referred to as the "far-end talker's single talk"), a switching operation is made such that the transmission of the near-end voice output Sout is prohibited or an operation is made such that the near-end voice output Sout is replaced by a pseudo noise.
  • the control unit 4 controls the echo path estimation/echo replica generation circuit 3 and the non-linear processor 6. That is, the control unit 4 detects the far-end taker's speechless condition or detects the double talk, controls the ON/OFF state of the learning function of the echo path estimation in accordance with a double talk signal DT, detects the far-end talker's single talk, and controls the switching operation of the non-linear processor 6.
  • the echo path estimation/echo replica generation circuit 3 estimates an echo path presuming that the far-end voice input Rin and the near-end voice input Sin are time-wise coincident with each other, and generates an echo based on the estimated echo path.
  • the near-end voice input Sin from the hybrid circuit 2 is delayed, with respect to the far-end voice input Rin, by a delay time attributable to a transmission path between the echo canceler 1 and the hybrid circuit 2, the far-end voice input Rin is input first in the echo path estimation/echo replica generation circuit 3 and then, the near-end voice input Sin corresponding to Rin is input therein with the above-mentioned delay time. During this time, it becomes impossible to satisfactorily carry out a learning based on the estimation of an echo path.
  • the echo canceler did not have any information of the echo path at the start of an operation.
  • characteristic of an echo path is substantially controlled by characteristic of a hybrid.
  • the waveform of an echo determined by a hybrid was longitudinally shifted on the time axis in accordance with a delay in the transmission line and attenuated in accordance with the attenuation in the transmission line.
  • a waveform of an echo in the transmission line with respect to an impulse input was obtained with a considerable accuracy.
  • a digital signal is transmitted in the transmission line and a D/A conversion (in general, ⁇ -LAW conversion) is performed between the echo canceler 1 for processing the digital signal and the hybrid circuit 2 for performing a conversion to an analog line. For this reason, a linear relation is established between the far-end voice output Rout and the near-end voice input Sin. Therefore, it is impossible to fully and completely cancel the echo merely by linear operation using the echo path estimation/echo replica generation circuit 3, etc.
  • the present invention has been accomplished in view of above-mentioned situation. It is, therefore, a first object of the invention to provide an echo canceler and an echo path estimating method capable of estimating an echo path rapidly and with a high accuracy.
  • a second object of the invention is to realize the first object without extensively altering the existing devices.
  • an echo canceler employed in a communication line network including a first transmission line for transmitting a voice of a four-wire side and a second transmission line for transmitting a voice of a second-wire side, the echo canceler comprising:
  • training signal generator means for generating a training signal and supplying the same to the first transmission line
  • coefficient calculator means for calculating a coefficient necessary for generating an echo replica based on a correlation established between the training signal supplied to the first transmission line and a signal of the second transmission line.
  • the training signal there can be listed a pseudo noise or a filtered pseudo noise.
  • the training signal is not particularly limited to one which is designed as a training signal, and a ring back tone or the like can be used for it.
  • a certain pseudo noise is supplied to a first transmission line for transmitting the far-end talker's voice and then a coefficient necessary for generating an echo replica is calculated based on a correlation established between the pseudo noise and a signal of a second transmission line.
  • This correlation is established under the condition that the voice level of the far-end talker's can be almost neglected.
  • the expression "such a condition that . . . can be almost neglected" used herein refers to the "time for calculating a correlation being long enough to sufficiently lower a correlation between a noise to be added and a voice on the four-wire side".
  • a training for estimating an echo path namely, a training for generating an echo replica can be performed irrespective of the far-end talker's voice.
  • the pseudo noise is subjected to filtering depending on the level or frequency characteristic of the near-end talker's voice, so that the characteristic may be varied. The reason is that even if a component of the pseudo noise is transmitted to the near-end talker through a hybrid circuit, no speech hindrance can be created. Further, since an echo replica is generated by adding a component of a pseudo noise, the component of the pseudo noise is eventually canceled and never transmitted to the far-end talker. Furthermore, if the voice signal corresponding to the voice on the four-wire side is delayed by a transmission delay occurred between the echo canceler and the hybrid circuit, the delayed first voice signal can be time-wise matched with the second voice signal corresponding to the voice on the two-wire side. Thus, there is almost no time for generating an echo replica only by the first voice signal. Since this specific improvement can be applied directly to the existing devices, the afore-mentioned entire modification, etc. of the echo canceler are not required.
  • the first and second voice signals are up-sampled, there can be obtained an echo replica with a high accuracy. Since this specific improvement can be applied directly to the existing devices, the entire modification, etc. of the echo canceler are not required.
  • the learning speed can be increased.
  • this is more preferable.
  • FIG. 1 is a block diagram showing a construction of a conventional echo canceler.
  • FIG. 2 is a chart showing the waveform of an echo.
  • FIG. 3 is a block diagram showing an important portion of an echo canceler according to a first embodiment of the present invention.
  • FIG. 4 is a block diagram showing an important portion of an echo canceler according to a second embodiment of the present invention.
  • FIG. 5A, 5B, and 5C are graphs showing frequency characteristics of a filter.
  • FIG. 6 is a block diagram showing an important portion of an echo canceler according to a third embodiment of the present invention.
  • FIG. 7 is a block diagram showing an important portion of an echo canceler according to a fourth embodiment of the present invention.
  • FIG. 8 is a block diagram showing an important portion of an echo canceler according to a modification of the fourth embodiment
  • FIG. 9 is a block diagram showing an important portion of an echo canceler according to a fifth embodiment of the present invention.
  • FIG. 10 is a block diagram showing an important portion of an echo canceler according to a sixth embodiment of the present invention.
  • FIG. 11 is a block diagram showing an important portion of an echo canceler according to a modified embodiment consisting of a combination of the third embodiment and the fourth embodiment.
  • FIG. 3 is a block diagram showing an important portion of an echo canceler according to the present invention.
  • a pseudo noise generator 11 generates a certain pseudo noise and outputs it.
  • a noise of a certain level for example, white noise
  • This pseudo noise is added to a far-end voice input Rin by an adder 14, and its result is output as a far-end voice output Rout. Therefore, a part of the pseudo noise is mixed with the near-end voice input Sin through the hybrid circuit 2 and supplied to a coefficient calculator 15.
  • the coefficient calculator 15 calculates a coefficient (for example, a tap coefficient such as a digital filter) necessary for generating an echo replica based on the far-end voice output Rout and the near-end voice input Sin.
  • a coefficient for example, a tap coefficient such as a digital filter
  • a pseudo noise to be added (or applied) to Rout is represented by n(t) and a signal to be obtained in Sin is represented by S(t).
  • h(t) is an impulse response of an echo.
  • N is set as shown by the following equation (4). ##EQU4##
  • the above h(t) is an impulse response of an echo and therefore, equal to the coefficient for generating an echo replica. This can be derived from the correlation shown by the above equation (2) as previously mentioned.
  • the coefficient calculator 15 calculates this coefficient h(t) by means of the above-mentioned calculation procedure and outputs the same to an echo replica generator 16.
  • This echo replica generator 16 generates an echo replica based on this coefficient. The details will now be described. Firstly, it is designed such that the echo replica generator 16, as in the case with a known adaptive filter, outputs an echo replica ya, based on the following equation (8).
  • x (x k-1 , X k-2 , . . . , x k-n ) t
  • x j x(JT)
  • T is a sampling interval
  • x(hT) is a sampling result of the far-end voice signal Rout at time jT)
  • the coefficients h 1 , h 2 , . . . , h n are set to h(T), h(2T), . . . , h(nT), respectively. Accordingly, an echo component contained in the near-end voice input Sin is canceled by an adder 17. Since such an echo replica is generated by, as previously mentioned, adding a pseudo noise which is mixed with the far-end voice output Rout, it can be canceled even in the event that the component of the pseudo noise is mixed with the near-end voice input Sin. As a consequence, it can be avoided that the component of the pseudo noise is transmitted to the far-end talker.
  • FIG. 4 is a block diagram showing an important portion of an echo canceler according to a second embodiment of the present invention.
  • a pseudo noise generator 11 as in the case with the comparable one of the first embodiment, generates a certain pseudo noise and outputs it.
  • a level/frequency characteristic measuring unit 12 measures the level and frequency characteristic of a signal of the near-end voice input Sin. Depending on a result of this measurement, characteristics of filters 13 and 18 are varied.
  • FIG. 5(a) shows a frequency characteristic of the above pseudo noise.
  • a pseudo noise having a flat characteristic is employed.
  • FIG. 5(b) shows a frequency characteristic of the near-end voice input Sin corresponding to the near-end talker's voice measured by the level/frequency characteristic measuring unit 12.
  • the characteristic of the filter 13 is varied as shown in FIG. 5(c) in accordance with the measured frequency characteristic.
  • the filter characteristic is variably set such that the frequency of the near-end talker's voice is simulated and the difference in level is fixed (in the illustrated example, 20dB).
  • the filter 18 is set such that it has an inverse characteristic with respect to that of the filter 13. With this feature, if the filters 13 and 18 are cascaded to each other, input and output signals of the cascaded circuit become equal to each other.
  • the filter 13 renders the variable filter characteristic thus set to pseudo noise and then outputs the noise. Consequently, frequency characteristic of the pseudo noise is varied in accordance with the near-end talker's voice. Since the characteristic of the pseudo noise, when output, will come to correspond to the near-end talker's voice, any adverse effects to the near-end talker caused by the pseudo noise can be avoided even if such a pseudo noise is admixed with the far-end talker's voice output Rout and transmitted to the near-end talker through the hybrid circuit 2.
  • An output from the filter 13 is supplied to the transmission line for transmitting the far-end talker's voice through the adder 14 and served as the afore-mentioned far-end voice output Rout. Therefore, the output from the filter 13 is partly admixed with the near-end voice input Sin through the hybrid circuit 2 and then supplied to the filter 18. Since the filter 18 has an inverse characteristic with respect to that of the filter 13, an output from the filter 18 becomes similar to a signal obtainable when a pseudo noise output from the pseudo noise generator 11 is supplied directly to the hybrid circuit 2. Next, the coefficient calculator 15 calculates a coefficient (for example, a tap coefficient of a digital filter or the like) necessary for generating an echo replica based on the far-end voice output Rout and near-end voice input Sin.
  • a coefficient for example, a tap coefficient of a digital filter or the like
  • frequency characteristic S(f) of the near-end voice signal Sin can be given by the following equation (9).
  • an impulse response of an echo namely, a coefficient for generating an echo replica can be obtained from a correlation computation (the under-listed equation (14)) between an output S (t) of the filter 18 and an output n (t) of the noise generator 11. ##EQU10##
  • the echo replica generator 16 as in the case with the known adaptive filter, outputs an echo replica ya based on the following equation (15) (as in the case with the equation (7) of the first embodiment).
  • the coefficients h 1 , h 2 , . . . , h n are set to ha (T), ha (2T), . . . , ha (nT), respectively. Therefore, the echo component contained in the near-end voice input Sin is canceled by the adder 17.
  • Such an echo replica is generated by adding a pseudo noise which is admixed with the far-end voice output Rout. Therefore, even if the component of the specific pseudo noise is admixed with the near-end voice input Sin, the noise can be canceled. Eventually, it can be avoided that the component of the pseudo noise is transmitted to the far-end talker.
  • the pseudo noise whose frequency characteristic is variable in accordance with the near-end talker's voice, is forcibly supplied to the transmission line for transmitting the far-end talker's voice, and an echo path is estimated and an echo replica is generated by using the specific pseudo noise. Accordingly, a training for estimating an echo path can be carried out irrespective of the far-end talker's voice.
  • an appropriate echo replica can be generated while minimizing a possible deterioration of speech quality of the near-end talker.
  • the far-end voice input Rin and the near-end voice input Sin are time-wise matched, so that accuracy of cancellation characteristic of the echo canceler is improved.
  • FIG. 6 is a block diagram showing a construction of the third embodiment.
  • An echo canceler 10 shown in this illustration is different from the conventional echo canceler 1 shown in FIG. 1 in the respect that a delay circuit 31 for delaying the far-end voice input Rin is located on a front stage of an echo path estimation/echo replica generation circuit 3.
  • a delay time in this delay circuit 31 is generally equal to a transmission delay occurrable between an echo canceler 10 and a hybrid circuit 2.
  • the near-end voice input Sin from the hybrid circuit 2 is delayed in response by the above-mentioned delay time with respect to the far-end voice input Rin. Accordingly, the far-end voice input Rin is input first in the echo path estimation/echo replica generation circuit 3 and then the corresponding near-end voice input Sin is input therein with the above-mentioned delayed time. For this reason, if the delay time corresponds to the calculating time of, for example, 200 taps of the FIR filter, it is only 312 taps among so many taps as 512 that can be substantially contributed to the estimation of an echo path and/or generation of an echo replica in the procedure of an echo path estimation/echo replica generation.
  • the delay time with respect to the hybrid circuit 2 is given to the far-end voice input Rin so that the far-end voice input Rin is time-wise matched with the near-end voice input Sin, the performance of the FIR filter can be fully extracted and an echo replica can be generated with a higher accuracy.
  • accuracy of the cancellation characteristic of the echo canceler can be improved.
  • the existing devices can be directly used in this embodiment, the afore-mentioned entire modification, etc., are no more required.
  • the far-end voice input Rin and the near-end voice input Sin are so-called over-sampled in order to improve accuracy of the cancellation characteristic of the echo canceler.
  • FIG. 7 is a block diagram showing a construction of the fourth embodiment.
  • An echo canceler 20 shown in this illustration is different from the echo canceler 1 shown in FIG. 1 in the respect that an up-sampler 32 for up-sampling the far-end voice input Rin and another up-sampler 33 for up-sampling the near-end voice input Sin are located at a front stage of an echo path estimation/echo replica generation circuit 3, and a down-sampler 34 for down-sampling an echo replica is located at a rear stage of the echo path estimation/echo replica generation circuit 3.
  • the sampling frequency of the up-samplers 32, 33 is 16 k Hz of 2 Fs.
  • the sampling frequency by the down-sampler 34 is 8 k Hz of Fs.
  • the estimation of an echo path and/or generation of an echo replica are performed at the same 2 Fs as the up-samplers 32, 33. Since this echo is generated by convolutional operation between an echo path estimated based on the up-sampled far-end voice input Rin and near-end voice input Sin and the up-sampled far-end voice input Rin, its accuracy is higher than the conventional echo canceler. As a consequence, accuracy of the cancellation characteristic of the echo canceler can be improved. Furthermore, there is such an advantage that the speed of convergence becomes faster as the sampling rate is increased. In addition, since the echo canceler of this embodiment, as in the case with the third embodiment, can use the existing devices as they are, the afore-mentioned entire modification, etc. are no more required.
  • FIG. 9 is a block diagram of an important portion of an echo canceler according to a fifth embodiment of the present invention.
  • an initial estimation circuit 7 is employed in addition to the echo canceler of FIG. 1.
  • a portion which is not illustrated is constructed in the same way as FIG. 1.
  • reference numeral 71 denotes a delay time measuring circuit.
  • the delay time measuring circuit 71 obtains a delay time T D (see FIG. 2) based on the near-end voice input Sin and the far-end voice output Rout and outputs its result.
  • T D a difference between the time when a peak of the near-end voice input Sin is generated and the time when a peak of the far-end voice output Rout is generated, both in a predetermined period may be obtained, or a mutual correlation between the two signals may be obtained.
  • Reference numeral 75 denotes a delay circuit. This delay circuit 75 delays the far-end voice output Rout by the delay time T D which is output from the delay time measuring circuit 71.
  • Reference numeral 72 denotes a level ratio measuring circuit. This level ratio measuring circuit 72 compares a level between the near-end voice input Sin and the far-end voice output Rout and supplies a signal corresponding to the ratio to one end of a multiplier 76. Also, the delayed far-end voice output Rout is supplied to another end of the multiplier 76. Accordingly, a far-end voice output Rout, which is normalized in delay time and level (this output is hereinafter referred to as "the signal Rout'"), is output from the multiplier 76.
  • Reference numeral 74 denotes a waveform data base.
  • This waveform data base 74 stores characteristic data of echo waveforms for various kinds of hybrids.
  • the terms "characteristic data” used herein refers to the tap coefficient b i (see the equation (1)) in the FIR filter of the echo path estimation/echo replica generation circuit 3. It should be noted that there are about dozens kinds of hybrids which are currently used in a telephone line and it is not a difficult job to store the characteristic data for all kinds of hybrids.
  • Reference numeral 73 denotes a configuration comparator. This configuration comparator 73 selects one of the characteristic data, which is most resemble the actual correlation, based on the correlation between the signal Rout' and the near-end voice input Sin. The selected characteristic data (tap coefficient b i ) is set to an initial value of the tap coefficient of the FIR filter of the echo path estimation/echo replica generation circuit 3.
  • the far-end voice output Rout which was delayed by the delay circuit 75, is supplied also to the echo path estimation/echo replica generation circuit 3.
  • no delay time T D exists between the near-end voice input Sin and the far-end voice output Rout.
  • the delay time measuring circuit 71, level ratio measuring circuit 72 and configuration comparator 73 are all stopped in operation when a double talk is detected in the control unit 4 (i.e., when the double talk signal DT is output).
  • the delay time measuring circuit 71 when the far-end talker issues some voice, its content is supplied to the delay time measuring circuit 71 in the form of the far-end voice output Rout.
  • the delay time T D is calculated in the delay time measuring circuit 71 and its value is supplied to the delay circuit 75 and the echo path estimation/echo replica generation circuit 3.
  • the far-end voice output Rout which is now delayed by T D , is supplied to the level ratio measuring circuit 72.
  • the level ratio measuring circuit 72 supplies a signal corresponding to a ratio of level between the near-end voice input Sin and the far-end voice output Rout to one end of the multiplier 76.
  • the signal Rout' which is now normalized in delay time and level, is supplied to the configuration comparator 73.
  • one of the characteristic data stored in the waveform data base 74 which is most resemble the actual correlation, is selected based on the correlation between the signal Rout' and the near-end voice input Sin.
  • the selected characteristic data is corrected its level by the multiplier 77 and is then supplied to the echo path estimation/echo replica generation circuit 3.
  • an initial value of the tap coefficient of the FIR filter is set.
  • the far-end voice output Rout which is now delayed by T D , is further supplied to the echo path estimation/echo replica generation circuit 3 through the delay circuit 75.
  • the operation of the initial estimation circuit 7 is finished. Thereafter, in the echo path estimation/echo replica generation circuit 3, a learning identification i s made based on the near-end voice input Sin and the far-end voice output Rout, and the tap coefficient b i is appropriately changed in order to estimate a more correct echo path.
  • the delay time of the delay circuit 75 can be set by the delay time measuring circuit 71 and an initial value of the tap coefficient b i in the echo path estimation/echo replica generation circuit 3 can be set by the configuration comparator 73 and the waveform data base 74.
  • the estimation of an echo path is not finished by those initial values and a further learning is required in the echo path estimation/echo replica generation circuit 3.
  • the time required for converging the result of learning can be remarkably shortened by rendering the initial values having a somewhat satisfactory degree of correctness to the echo path estimation/echo replica generation circuit 3.
  • the far-end voice output Rout which is now delayed by T D , is supplied to the echo path estimation/echo replica generation circuit 3 through the delay circuit 75, load of the learning in the echo path estimation/echo replica generation circuit 3 can be reduced and an echo path estimation can be performed at a higher speed and with a higher accuracy.
  • an echo path estimation can be performed at a high speed and with a high accuracy.
  • FIG. 10 is a block diagram of an important portion of an echo canceler according to one embodiment of the present invention.
  • a spectrum comparator 81 and a frequency response data base 81 are employed in the place of the configuration comparator 73 and the waveform data base 74 of the fifth embodiment.
  • the frequency response data base 81 stores data of frequency response characteristics of various kinds of hybrids. Since there are only about dozens kinds of hybrids which are currently used in a telephone line as previously mentioned, it is not a difficult job to store data of the frequency response characteristics of all kinds of hybrids.
  • the spectrum comparator 81 selects one of the frequency response characteristics, which is most resemble the actual characteristic, based on the correlation between the signal Rout' and the near-end voice input Sin.
  • a convertor 83 converts the selected frequency response characteristic to a response characteristic on a time axis. Specifically, the frequency response characteristics are converted to the tap coefficients b i and those tap coefficients b i are set to initial values of the tap coefficients of the FIR filer of the echo path estimation/echo replica generation circuit 3.
  • the spectrum comparator 81 is, as in the case with the configuration comparator 73 of the fifth embodiment, stopped in operation when a double talk is detected (i.e., when a double talk signal DT is output) in the control unit 4.
  • the delay time measuring circuit 71 when the far-end talker issues some voice, its content is supplied to the delay time measuring circuit 71 in the form of the far-end voice output Rout.
  • the delay time T D is calculated in the delay time measuring circuit 71 and its value is supplied to the delay circuit 75 and the echo path estimation/echo replica generation circuit 3.
  • the far-end voice output Rout which is now delayed by T D , is supplied to the level ratio measuring circuit 72.
  • the level ratio measuring circuit 72 supplied a signal corresponding to a ratio of level between the near-end voice input Sin and the far-end voice output Rout to one end of the multiplier 76.
  • the signal Rout' which is now normalized in delay time and level, is supplied to the spectrum comparator 81.
  • one of the characteristic data stored in the frequency response data base 82 which is most resemble the actual characteristic, is selected based on the correlation between the signal Rout' and the near-end voice input Sin.
  • the selected characteristic is converted to the tap coefficient b i through the convertor 83, then corrected its level by the multiplier 77 and then supplied to the echo path estimation/echo replica generation circuit 3.
  • an initial value of the tap coefficient of the FIR filter is set.
  • the operation of the initial estimation circuit 7 is finished. Thereafter, in the echo path estimation/echo replica generation circuit 3, a learning identification is made based on the near-end voice input Sin and the far-end voice output Rout, and the tap coefficient b i is appropriately changed in order to estimate a more correct echo path.
  • the delay time of the delay circuit 75 can be set by the delay time measuring circuit 71 and an initial value of the tap coefficient b i can be set by the delay time measuring circuit 71, frequency response data base 81 and convertor 83.
  • the time required for converging the result of learning can be remarkably shortened as in the case with the fifth embodiment by rendering the initial values having a somewhat satisfactory degree of correctness to the echo path estimation/echo replica generation circuit 3.
  • the present invention is applied to a signal transmission between a mobile telephone and a terrestrial telephone.
  • application of the present invention is not limited to this.
  • the present invention can be applied to all communication networks in which a signal transmission is made between a two-wire circuit and a four-wire circuit.
  • the pseudo noise generator 11 normally generates a pseudo noise.
  • an arrangement may also be made, in which a call boundary is received from the switch and no pseudo signal is generated until after the passage of a certain time. Owing to this arrangement, the learning can be finished by the time the near-end talker brings the receiver very close to the ear, so that a pseudo noise can be prevented from being transmitted to the near-end talker at the time a speech is actually made.
  • the pseudo noise generator 11 may generate a pseudo noise when the far-end talker remains in a speechless condition for more than a predetermined time. The reason is that when the far-end talker side is in a speechless condition, transmission of a noise, if any, to the near-end talker side cannot be any hindrance to the speech.
  • the pseudo noise generator 11 may generate a pseudo noise when a cancellation of echo has gone below a predetermined level in the adder 17. The reason is that there is a high possibility of no generation of normal echo replica in the echo replica generator 16 and a re-learning seems to be desirable.
  • the level of a pseudo noise is constant in the first embodiment. However, the level of a pseudo noise may be decreased as the time required for carrying out the correlation operation is increased.
  • the up-samplers 32, 33 and the down-sampler 34 are constituted of separate elements.
  • they may be constituted of a single element by a digital signal processor (DSP) 35 as shown in FIG. 8.
  • DSP digital signal processor
  • the order of the sampling procedure may be changed in accordance with necessity. Important things here are that the processing timing of the up-samplers 32, 33 and the down-sampler 34 must be time-wise coincident and the output from the down-sampler 34 must be time-wise continuous.
  • FIG. 11 shows one example of a construction of an echo canceler 30 which is comprised of such a combination.
  • a delay circuit 31 is located at a front stage of the up-sampler 32.
  • This delay circuit 31 is adapted to delay the far-end voice input Rin by a transmission delay occurrable between the echo canceler 30 and the hybrid circuit 2.
  • accuracy of the cancellation characteristic of the echo canceler can be more improved by the multiplier effect of the third and fourth embodiments.
  • the waveform data base 74 stores the tap coefficient b i as the characteristic data.
  • the characteristic data is not limited to the tap coefficient b i .
  • an alpha parameter, an LSP parameter, a per-call coefficient or the like which can indicate the response characteristics of various kinds of hybrids, may be stored as the characteristic data.
  • the echo path estimation/echo replica generation circuit 3 is not limited to one which employs the learning identification algorithm. In the alternative, a wide range of algorithms such as Kalman filter and the like may be employed.
  • the echo path estimation/echo replica generation circuit 3 may be redesigned such that at first, the level of the echo replica is set to be smaller than the estimated value and it is gradually increased to the original level with the passage of time. The reason is that since it is difficult to estimate an echo replica correctly at the early stage, a high setting of the level of the echo replica signal Y rather results in such an inconvenience as a generation of a noise, etc.
  • an echo path is estimated based on a voice produced by the far-end talker.
  • an echo path may be estimated using other signals. For example, in the case where a calling is issued from the near-end, an echo path may be estimated using a ring back tone, whereas in the case where a calling is issued from the far-end, an echo path may be estimated by transmitting a training signal to the hybrid immediately after the off-hook is made on the near-end side so that an echo path is estimated based thereon. The reason is that owing to the foregoing arrangement, when the near-end talker or the far-end talker actually starts a speech, an accurate echo path estimation is already obtained.
  • the far-end voice output Rout which has been delayed by the delay circuit 75, is supplied to the echo path estimation/echo replica generation circuit 3.
  • a non-delayed far-end voice output Rout may be supplied to the echo path estimation/echo replica generation circuit 3.
  • the far-end voice output Rout may be delayed by the delay time T D measured in the delay time measuring circuit 71, within the echo path estimation/echo replica generation circuit 3 (specifically, in a numerical FIG. 1, a corresponding tap coefficient b i is brought to "0").

Abstract

The present invention relates to an echo canceler for training an echo path estimation without providing a hindrance to a speech. The echo canceler of the present invention includes a pseudo noise generator for generating a certain pseudo noise. This pseudo noise is forcibly supplied to a transmission line for transmitting a far-end talker's voice. Here, in the case where the far-end talker's transmitting speech level is faint or the far-end talker is in a speechless condition, a certain correlation is established between the pseudo noise and a signal of the transmission line for transmitting the near-end talker's voice. Based on such a correlation as just mentioned, a coefficient for generating an echo replica is calculated. Therefore, a training for estimating an echo path is performed based on the pseudo noise and without depending on the far-end talker's voice, thereby enabling to generate an appropriate echo replica.

Description

TECHNICAL FIELD
This invention relates to an echo canceler suited to be used for a mobile communications network and a long-distance telephone line network. Also, the invention relates to an echo path estimating method suited to be used for this echo canceler.
BACKGROUND ART
In a long-distance telephone line via a submarine cable or via a communication satellite, the subscriber's line, in general, connected to both ends of the line is of a two-wire circuit and its long-distance transmission portion is of a four-wire circuit for the purposes of amplification of a signal, etc. Similarly, in the mobile communications network using a mobile telephone (or cellular phone), the subscriber's line of a terrestrial analog telephone is of a two-wire circuit and its portion from a terminal of the mobile telephone to a switch, etc. is of a four-wire circuit. In this case, the connection region between the two-wire and the four-wire is provided with a hybrid circuit for performing a four-wire/two-wire conversion.
This hybrid circuit is designed to match with the impedance of the two-wire circuit. However, since it is difficult to obtain always a good matching condition, a received signal reaching an input side of the four-wire of the hybrid circuit tends to leak toward an output side of the four-wire, thereby generating a so-called echo. Since such an echo is lower in level than the talker's voice and reaches the talker after a delay of a predetermined time period, a speech hindrance is created. Such a speech hindrance caused by echo becomes more significant as the signal propagation time becomes longer. Particularly, in the case of a mobile communication using a mobile telephone, since various processing procedures are carried out in the radio communication section leading to the switch, etc., the delay of signal is increased, thus resulting, particularly, in the problem of speech hindrance caused by echo. FIG. 2 shows one example of the waveform of an echo with respect to impulse response.
As an apparatus for preventing a generation of echo, there are an echo suppressor and an echo canceler. FIG. 1 shows a schematic construction of an echo canceler which can be used in a mobile communications network. The echo canceler 1 illustrated here is located on a front stage of a hybrid circuit 2. In this illustration, the subscriber of an ordinary analog telephone is referred to as the "near-end talker" and the subscriber of a mobile telephone as the "far-end talker". A far-end voice signal input into the echo canceler 1 is represented by Rin; a far-end voice signal output from the echo canceler 1, by Rout; a near-end voice signal input into the echo canceler 1, by Sin; and a near-end voice signal output from the echo canceler 1, by Sout; respectively.
The echo canceler 1 shown in FIG. 1 comprises an echo path estimation/echo replica generation circuit 3, a control unit 4, an adder 5, and a non-linear processor 6. Here, the echo path estimation/echo replica generation circuit 3 detects a response characteristic of the hybrid circuit 2 based on both the far-end voice input Rin and near-end voice input Sin and estimates an echo path (namely, echo propagating line). Then, an anticipated echo (namely, echo replica) from the hybrid circuit 2 is generated through a convolutional operation based on the result of estimation and the far-end voice input Rin. This echo replica is generated by an FIR filter which is constituted of so many taps as 512, for example. A convolutional operation in an echo replica refers to this. In the adder 5, this echo replica is subtracted from the near-end voice input Sin, thereby canceling the echo. As the above-mentioned echo path estimation algorithm, a learning identification algorithm is used. Among many adaptive algorithms, this learning identification algorithm is comparatively small in computational complexity and good in convergence characteristic.
Specifically, the echo path estimation/echo replica generation circuit 3 has an FIR filter. An echo replica signal Y(z) output from the FIR filter can be obtained by the following equation (1). ##EQU1##
In the equation (1), N is the number of taps of the FIR filer, and bi (where i=0, 1, 2, . . . N-1) is a tap coefficient in each tap. If appropriate values of the tap number N and tap coefficient bi can be obtained by estimation of an echo path, the echo replica signal Y(z) is approximated to an actual echo. Thus, echo is canceled in the adder 5. As the above-mentioned echo path estimation algorithm, an adaptive filter technique, for example, a learning identification algorithm, which among many other adaptive algorithms, is comparatively small in computational complexity and good in convergence characteristic, is used. Details of the learning identification algorithm is disclosed, for example, in Institute of Electronics and Communication Engineers of Japan (IECE) Journal '77/11 Vol. J60-A NO.11, article under the heading of "Regarding Echo Canceling Characteristic of Echo Canceler Using Learning Identification Algorithm".
As conditions for enabling the above learning, the following requirements must be met.
1 A far-end voice output Rout of the level sufficient for an echo to come back as a near-end voice input Sin is present. In other words, the far-end taker is currently engaged in speech.
2 The near-end voice input Sin is constituted of an echo (or an echo and a white noise) alone. In other words, the near-end taker is not engaged in speech.
On the other hand, when the far-end talker is in a speechless condition and when the far-end talker and the near-end talker are simultaneously engaged in speech (this state is hereinafter referred to as the "double talk"), it is necessary to turn off the learning function because there is a fear to cause a mis-learning state of echo path estimation.
In the transmission line, digital signals are transmitted, and a D/A conversion (in a general expression, a μ-LAW conversion) is made between the echo canceler 1 adapted to process such digital signals and the hybrid circuit 2 adapted to undertake a conversion to the analog line. For this reason, a non-linear characteristic relation is established between the far-end voice output Rout and the near-end voice input Sin. Therefore, echo cannot be canceled fully and completely merely through the linear computation by means of the echo path estimation/echo replica generation circuit 3, etc. As a consequence, an echo component unable to be completely canceled is produced.
In order to remove such an echo component (hereinafter referred to as the "residual echo"), the non-linear processor 6 is employed. This non-linear processor 6 undertakes a non-linear switching operation. Specifically, in the case where the near-end voice output Sout is constituted merely of an echo, in other words, in the case where only the far-end talker is currently engaged in speech (this state is hereinafter referred to as the "far-end talker's single talk"), a switching operation is made such that the transmission of the near-end voice output Sout is prohibited or an operation is made such that the near-end voice output Sout is replaced by a pseudo noise.
The control unit 4 controls the echo path estimation/echo replica generation circuit 3 and the non-linear processor 6. That is, the control unit 4 detects the far-end taker's speechless condition or detects the double talk, controls the ON/OFF state of the learning function of the echo path estimation in accordance with a double talk signal DT, detects the far-end talker's single talk, and controls the switching operation of the non-linear processor 6.
Incidentally, in the above-mentioned techniques, there are encountered with the following problems.
1 Firstly, since the above-mentioned techniques merely employ an adaptive filter technology such as the learning identification algorithm, if the delay time of an echo to be canceled is increased, the number of taps of the adaptive filter is increased and the computational complexity is also increased.
In other words, the echo path estimation/echo replica generation circuit 3 estimates an echo path presuming that the far-end voice input Rin and the near-end voice input Sin are time-wise coincident with each other, and generates an echo based on the estimated echo path. However, since the near-end voice input Sin from the hybrid circuit 2 is delayed, with respect to the far-end voice input Rin, by a delay time attributable to a transmission path between the echo canceler 1 and the hybrid circuit 2, the far-end voice input Rin is input first in the echo path estimation/echo replica generation circuit 3 and then, the near-end voice input Sin corresponding to Rin is input therein with the above-mentioned delay time. During this time, it becomes impossible to satisfactorily carry out a learning based on the estimation of an echo path.
2 Also, in the above-mentioned conventional techniques, the echo canceler did not have any information of the echo path at the start of an operation. However, observation of the present inventors revealed that characteristic of an echo path is substantially controlled by characteristic of a hybrid. Specifically, the waveform of an echo determined by a hybrid was longitudinally shifted on the time axis in accordance with a delay in the transmission line and attenuated in accordance with the attenuation in the transmission line. As a result, a waveform of an echo in the transmission line with respect to an impulse input was obtained with a considerable accuracy.
3 A digital signal is transmitted in the transmission line and a D/A conversion (in general, μ-LAW conversion) is performed between the echo canceler 1 for processing the digital signal and the hybrid circuit 2 for performing a conversion to an analog line. For this reason, a linear relation is established between the far-end voice output Rout and the near-end voice input Sin. Therefore, it is impossible to fully and completely cancel the echo merely by linear operation using the echo path estimation/echo replica generation circuit 3, etc.
In order to improve the shortcomings of 1 to 3 as a group, it is necessary to change the design of the echo canceler extensively or to modify it entirely. This being the case, it was unexpectable that the existing equipment is effectively used. Further, recently, there is a demand of a high-speed convergence with respect to an echo canceler.
DISCLOSURE OF INVENTION
The present invention has been accomplished in view of above-mentioned situation. It is, therefore, a first object of the invention to provide an echo canceler and an echo path estimating method capable of estimating an echo path rapidly and with a high accuracy.
A second object of the invention is to realize the first object without extensively altering the existing devices.
To achieve the above objects, according to the present invention, there is provided an echo canceler employed in a communication line network including a first transmission line for transmitting a voice of a four-wire side and a second transmission line for transmitting a voice of a second-wire side, the echo canceler comprising:
training signal generator means for generating a training signal and supplying the same to the first transmission line; and
coefficient calculator means for calculating a coefficient necessary for generating an echo replica based on a correlation established between the training signal supplied to the first transmission line and a signal of the second transmission line.
Here, as one example of the training signal, there can be listed a pseudo noise or a filtered pseudo noise. The training signal is not particularly limited to one which is designed as a training signal, and a ring back tone or the like can be used for it.
In the case where a pseudo noise is selected as the training signal, a certain pseudo noise is supplied to a first transmission line for transmitting the far-end talker's voice and then a coefficient necessary for generating an echo replica is calculated based on a correlation established between the pseudo noise and a signal of a second transmission line. This correlation is established under the condition that the voice level of the far-end talker's can be almost neglected. The expression "such a condition that . . . can be almost neglected" used herein refers to the "time for calculating a correlation being long enough to sufficiently lower a correlation between a noise to be added and a voice on the four-wire side". In other words, a training for estimating an echo path, namely, a training for generating an echo replica can be performed irrespective of the far-end talker's voice.
It is more preferred here that the pseudo noise is subjected to filtering depending on the level or frequency characteristic of the near-end talker's voice, so that the characteristic may be varied. The reason is that even if a component of the pseudo noise is transmitted to the near-end talker through a hybrid circuit, no speech hindrance can be created. Further, since an echo replica is generated by adding a component of a pseudo noise, the component of the pseudo noise is eventually canceled and never transmitted to the far-end talker. Furthermore, if the voice signal corresponding to the voice on the four-wire side is delayed by a transmission delay occurred between the echo canceler and the hybrid circuit, the delayed first voice signal can be time-wise matched with the second voice signal corresponding to the voice on the two-wire side. Thus, there is almost no time for generating an echo replica only by the first voice signal. Since this specific improvement can be applied directly to the existing devices, the afore-mentioned entire modification, etc. of the echo canceler are not required.
Moreover, if the first and second voice signals are up-sampled, there can be obtained an echo replica with a high accuracy. Since this specific improvement can be applied directly to the existing devices, the entire modification, etc. of the echo canceler are not required.
Furthermore, if it is designed such that response characteristics or frequency characteristics of a plurality of hybrids are stored in a storage unit, one of the characteristics is selected based on the correlation between the transmission signal transmitted from the four-wire side to the two-wire side and the echo signal transmitted from the two-wire side to the four-wire side, and the various parameters are initialized based on the selected characteristic, the learning speed can be increased. Thus, this is more preferable.
BRIEF DESCRIPTION OF DRAWINGS
FIG. 1 is a block diagram showing a construction of a conventional echo canceler.
FIG. 2 is a chart showing the waveform of an echo.
FIG. 3 is a block diagram showing an important portion of an echo canceler according to a first embodiment of the present invention.
FIG. 4 is a block diagram showing an important portion of an echo canceler according to a second embodiment of the present invention.
FIG. 5A, 5B, and 5C are graphs showing frequency characteristics of a filter.
FIG. 6 is a block diagram showing an important portion of an echo canceler according to a third embodiment of the present invention.
FIG. 7 is a block diagram showing an important portion of an echo canceler according to a fourth embodiment of the present invention;
FIG. 8 is a block diagram showing an important portion of an echo canceler according to a modification of the fourth embodiment;
FIG. 9 is a block diagram showing an important portion of an echo canceler according to a fifth embodiment of the present invention;
FIG. 10 is a block diagram showing an important portion of an echo canceler according to a sixth embodiment of the present invention; and
FIG. 11 is a block diagram showing an important portion of an echo canceler according to a modified embodiment consisting of a combination of the third embodiment and the fourth embodiment.
BEST MODE FOR CARRYING OUT THE INVENTION
(FIRST EMBODIMENT)
FIG. 3 is a block diagram showing an important portion of an echo canceler according to the present invention. Here, a pseudo noise generator 11 generates a certain pseudo noise and outputs it. As this pseudo noise, a noise of a certain level (for example, white noise) is used. This pseudo noise is added to a far-end voice input Rin by an adder 14, and its result is output as a far-end voice output Rout. Therefore, a part of the pseudo noise is mixed with the near-end voice input Sin through the hybrid circuit 2 and supplied to a coefficient calculator 15.
The coefficient calculator 15 calculates a coefficient (for example, a tap coefficient such as a digital filter) necessary for generating an echo replica based on the far-end voice output Rout and the near-end voice input Sin. In that case, a correlation established under the condition that the far-end talker's voice is faint or the far-end talker is in a speechless condition as shown by the following equation (2). ##EQU2##
In the above equation (2), t represents time and t=0 is time for starting a measurement. A pseudo noise to be added (or applied) to Rout is represented by n(t) and a signal to be obtained in Sin is represented by S(t). h(t) is an impulse response of an echo. Here, since n(t) is a noise which is close to white, a relation shown in the following equation (3) can be established with respect to a sufficiently large value TL. ##EQU3##
Here, δ(t) is a delta function and becomes 1 in case of t=0, and 0 in other cases. N is set as shown by the following equation (4). ##EQU4##
If the relation of the above equation (3) is used, ha(t), this being an estimated value of h(t), can be developed in a way as shown by the following equation (5). ##EQU5##
In the above equation (5), a portion shown by the following equation (6) becomes 1 in case of t=τ', and 0 in other cases. ##EQU6##
Therefore, the equation (5) can be approximated as shown by the equation (7). Eventually, the estimated value ha(t) becomes generally equal to h(t). ##EQU7##
It should be noted that the above h(t) is an impulse response of an echo and therefore, equal to the coefficient for generating an echo replica. This can be derived from the correlation shown by the above equation (2) as previously mentioned. The coefficient calculator 15 calculates this coefficient h(t) by means of the above-mentioned calculation procedure and outputs the same to an echo replica generator 16. This echo replica generator 16 generates an echo replica based on this coefficient. The details will now be described. Firstly, it is designed such that the echo replica generator 16, as in the case with a known adaptive filter, outputs an echo replica ya, based on the following equation (8).
ya=ha.sup.t x                                              (8)
where ha=(h1, h2, . . . , hn)t, (t is a transposition of a vector)
x=(xk-1, Xk-2, . . . , xk-n)t, xj =x(JT), (T is a sampling interval, and x(hT) is a sampling result of the far-end voice signal Rout at time jT)
In this embodiment, the coefficients h1, h2, . . . , hn are set to h(T), h(2T), . . . , h(nT), respectively. Accordingly, an echo component contained in the near-end voice input Sin is canceled by an adder 17. Since such an echo replica is generated by, as previously mentioned, adding a pseudo noise which is mixed with the far-end voice output Rout, it can be canceled even in the event that the component of the pseudo noise is mixed with the near-end voice input Sin. As a consequence, it can be avoided that the component of the pseudo noise is transmitted to the far-end talker. Therefore, a speech hindrance occurrable to t he far-end talker by the admixture of the pseudo noise is occurred. If a noise should be somehow transmitted to the near-end talker or far-end talker by the admixture of the pseudo noise, a possible speech hindrance would be avoided by appropriately adjusting the level of the specific pseudo noise, etc.
(SECOND EMBODIMENT)
FIG. 4 is a block diagram showing an important portion of an echo canceler according to a second embodiment of the present invention. In this embodiment, a pseudo noise generator 11, as in the case with the comparable one of the first embodiment, generates a certain pseudo noise and outputs it. On the other hand, a level/frequency characteristic measuring unit 12 measures the level and frequency characteristic of a signal of the near-end voice input Sin. Depending on a result of this measurement, characteristics of filters 13 and 18 are varied.
FIG. 5(a) shows a frequency characteristic of the above pseudo noise. As shown in this illustration, a pseudo noise having a flat characteristic is employed. FIG. 5(b) shows a frequency characteristic of the near-end voice input Sin corresponding to the near-end talker's voice measured by the level/frequency characteristic measuring unit 12. The characteristic of the filter 13 is varied as shown in FIG. 5(c) in accordance with the measured frequency characteristic. In this embodiment, the filter characteristic is variably set such that the frequency of the near-end talker's voice is simulated and the difference in level is fixed (in the illustrated example, 20dB). The filter 18 is set such that it has an inverse characteristic with respect to that of the filter 13. With this feature, if the filters 13 and 18 are cascaded to each other, input and output signals of the cascaded circuit become equal to each other.
The filter 13 renders the variable filter characteristic thus set to pseudo noise and then outputs the noise. Consequently, frequency characteristic of the pseudo noise is varied in accordance with the near-end talker's voice. Since the characteristic of the pseudo noise, when output, will come to correspond to the near-end talker's voice, any adverse effects to the near-end talker caused by the pseudo noise can be avoided even if such a pseudo noise is admixed with the far-end talker's voice output Rout and transmitted to the near-end talker through the hybrid circuit 2. The reason is that owing to the feature of the auditory sense, man hardly has a sense of physical disorder with respect to a signal whose frequency characteristic is approximated, and a possible deterioration in quality occurrable to a speech can be prevented from the viewpoint of man's physical sense of feel.
Also, it is set such that the higher the level of the far-end talker's voice becomes, the higher the gain of the filter 18 becomes. This arrangement is also for the same reason as mentioned above. Namely, when the talk level is high, the noise can be hardly recognized by man even if the noise level is comparatively high.
An output from the filter 13 is supplied to the transmission line for transmitting the far-end talker's voice through the adder 14 and served as the afore-mentioned far-end voice output Rout. Therefore, the output from the filter 13 is partly admixed with the near-end voice input Sin through the hybrid circuit 2 and then supplied to the filter 18. Since the filter 18 has an inverse characteristic with respect to that of the filter 13, an output from the filter 18 becomes similar to a signal obtainable when a pseudo noise output from the pseudo noise generator 11 is supplied directly to the hybrid circuit 2. Next, the coefficient calculator 15 calculates a coefficient (for example, a tap coefficient of a digital filter or the like) necessary for generating an echo replica based on the far-end voice output Rout and near-end voice input Sin. This principle will now be described in detail. Firstly, if a noise is represented by N(f) with respect to a frequency f; characteristic of the filter 11, by G(f); characteristic of the filter 18, by G-1 (f); and characteristic of an echo, by H(f), respectively, frequency characteristic S(f) of the near-end voice signal Sin can be given by the following equation (9).
S(f)=H(f)G(f)N(f)                                          (9)
Nextly, an output signal S'(f) of the filter 18 is given by the following equation (10). ##EQU8##
A correlation computation output between this and N (f) can be expressed by the following equation (11). ##EQU9##
In the above equation, since N (f) is a noise close to white, the following equation (12) is approximately established.
N(f)N(f)≈1                                         (12)
Therefore, the following equation (13) is established and an impulse response of an echo can be approximately obtained.
Ha(f)≈H(f)                                         (13)
Therefore, in a time area, an impulse response of an echo, namely, a coefficient for generating an echo replica can be obtained from a correlation computation (the under-listed equation (14)) between an output S (t) of the filter 18 and an output n (t) of the noise generator 11. ##EQU10##
The echo replica generator 16, as in the case with the known adaptive filter, outputs an echo replica ya based on the following equation (15) (as in the case with the equation (7) of the first embodiment).
ya=ha.sup.t x                                              (15)
where ha=(h1, h2, . . . ,hn)t
x=(xk-1, xk-2, . . . , xk-n)t, xj =x (jT)
In this embodiment, the coefficients h1, h2, . . . , hn, are set to ha (T), ha (2T), . . . , ha (nT), respectively. Therefore, the echo component contained in the near-end voice input Sin is canceled by the adder 17. Such an echo replica, as mentioned above, is generated by adding a pseudo noise which is admixed with the far-end voice output Rout. Therefore, even if the component of the specific pseudo noise is admixed with the near-end voice input Sin, the noise can be canceled. Eventually, it can be avoided that the component of the pseudo noise is transmitted to the far-end talker. Therefore, a speech hindrance occurrable to the far-end talker by the admixture of the pseudo noise is not occurred. If a noise should be somehow transmitted to the near-end talker or far-end talker by the admixture of the pseudo noise, such a possible speech hindrance would be avoided by appropriately adjusting the level of the specific pseudo noise, etc.
As mentioned above, according to this embodiment, the pseudo noise, whose frequency characteristic is variable in accordance with the near-end talker's voice, is forcibly supplied to the transmission line for transmitting the far-end talker's voice, and an echo path is estimated and an echo replica is generated by using the specific pseudo noise. Accordingly, a training for estimating an echo path can be carried out irrespective of the far-end talker's voice. Thus, by transfiguring the noise in accordance with the level/frequency of the near-end voice, an appropriate echo replica can be generated while minimizing a possible deterioration of speech quality of the near-end talker.
(THIRD EMBODIMENT)
A third embodiment of the present invention will now be described. In this embodiment, the far-end voice input Rin and the near-end voice input Sin are time-wise matched, so that accuracy of cancellation characteristic of the echo canceler is improved.
FIG. 6 is a block diagram showing a construction of the third embodiment. An echo canceler 10 shown in this illustration is different from the conventional echo canceler 1 shown in FIG. 1 in the respect that a delay circuit 31 for delaying the far-end voice input Rin is located on a front stage of an echo path estimation/echo replica generation circuit 3. A delay time in this delay circuit 31 is generally equal to a transmission delay occurrable between an echo canceler 10 and a hybrid circuit 2.
Next, effects of the above construction will be described. As mentioned above, in the prior art having no delay circuit 31, the near-end voice input Sin from the hybrid circuit 2 is delayed in response by the above-mentioned delay time with respect to the far-end voice input Rin. Accordingly, the far-end voice input Rin is input first in the echo path estimation/echo replica generation circuit 3 and then the corresponding near-end voice input Sin is input therein with the above-mentioned delayed time. For this reason, if the delay time corresponds to the calculating time of, for example, 200 taps of the FIR filter, it is only 312 taps among so many taps as 512 that can be substantially contributed to the estimation of an echo path and/or generation of an echo replica in the procedure of an echo path estimation/echo replica generation.
In contrast, in this embodiment, since the delay time with respect to the hybrid circuit 2 is given to the far-end voice input Rin so that the far-end voice input Rin is time-wise matched with the near-end voice input Sin, the performance of the FIR filter can be fully extracted and an echo replica can be generated with a higher accuracy. Thus, accuracy of the cancellation characteristic of the echo canceler can be improved. Moreover, since the existing devices can be directly used in this embodiment, the afore-mentioned entire modification, etc., are no more required.
(FOURTH EMBODIMENT)
A fourth embodiment of the present invention will now be described. In this embodiment, the far-end voice input Rin and the near-end voice input Sin are so-called over-sampled in order to improve accuracy of the cancellation characteristic of the echo canceler.
FIG. 7 is a block diagram showing a construction of the fourth embodiment. An echo canceler 20 shown in this illustration is different from the echo canceler 1 shown in FIG. 1 in the respect that an up-sampler 32 for up-sampling the far-end voice input Rin and another up-sampler 33 for up-sampling the near-end voice input Sin are located at a front stage of an echo path estimation/echo replica generation circuit 3, and a down-sampler 34 for down-sampling an echo replica is located at a rear stage of the echo path estimation/echo replica generation circuit 3.
Here, if, for example, the sampling frequency of the far-end voice input Rin and near-end voice input Sin is 8 k Hz (=Fs), the sampling frequency of the up- samplers 32, 33 is 16 k Hz of 2 Fs. Similarly, the sampling frequency by the down-sampler 34 is 8 k Hz of Fs.
Next, effects of the above-mentioned construction will be described. With such a construction as mentioned above, the estimation of an echo path and/or generation of an echo replica are performed at the same 2 Fs as the up- samplers 32, 33. Since this echo is generated by convolutional operation between an echo path estimated based on the up-sampled far-end voice input Rin and near-end voice input Sin and the up-sampled far-end voice input Rin, its accuracy is higher than the conventional echo canceler. As a consequence, accuracy of the cancellation characteristic of the echo canceler can be improved. Furthermore, there is such an advantage that the speed of convergence becomes faster as the sampling rate is increased. In addition, since the echo canceler of this embodiment, as in the case with the third embodiment, can use the existing devices as they are, the afore-mentioned entire modification, etc. are no more required.
(FIFTH EMBODIMENT)
A. Construction of the Embodiment
FIG. 9 is a block diagram of an important portion of an echo canceler according to a fifth embodiment of the present invention. In this device, an initial estimation circuit 7 is employed in addition to the echo canceler of FIG. 1. A portion which is not illustrated is constructed in the same way as FIG. 1.
In FIG. 9, reference numeral 71 denotes a delay time measuring circuit. The delay time measuring circuit 71 obtains a delay time TD (see FIG. 2) based on the near-end voice input Sin and the far-end voice output Rout and outputs its result. As means for obtaining the delay time TD, for example, a difference between the time when a peak of the near-end voice input Sin is generated and the time when a peak of the far-end voice output Rout is generated, both in a predetermined period may be obtained, or a mutual correlation between the two signals may be obtained.
Reference numeral 75 denotes a delay circuit. This delay circuit 75 delays the far-end voice output Rout by the delay time TD which is output from the delay time measuring circuit 71. Reference numeral 72 denotes a level ratio measuring circuit. This level ratio measuring circuit 72 compares a level between the near-end voice input Sin and the far-end voice output Rout and supplies a signal corresponding to the ratio to one end of a multiplier 76. Also, the delayed far-end voice output Rout is supplied to another end of the multiplier 76. Accordingly, a far-end voice output Rout, which is normalized in delay time and level (this output is hereinafter referred to as "the signal Rout'"), is output from the multiplier 76.
Reference numeral 74 denotes a waveform data base. This waveform data base 74 stores characteristic data of echo waveforms for various kinds of hybrids. The terms "characteristic data" used herein refers to the tap coefficient bi (see the equation (1)) in the FIR filter of the echo path estimation/echo replica generation circuit 3. It should be noted that there are about dozens kinds of hybrids which are currently used in a telephone line and it is not a difficult job to store the characteristic data for all kinds of hybrids. Reference numeral 73 denotes a configuration comparator. This configuration comparator 73 selects one of the characteristic data, which is most resemble the actual correlation, based on the correlation between the signal Rout' and the near-end voice input Sin. The selected characteristic data (tap coefficient bi) is set to an initial value of the tap coefficient of the FIR filter of the echo path estimation/echo replica generation circuit 3.
The far-end voice output Rout, which was delayed by the delay circuit 75, is supplied also to the echo path estimation/echo replica generation circuit 3. Thus, in the echo path estimation/echo replica generation circuit 3, no delay time TD (see FIG. 2) exists between the near-end voice input Sin and the far-end voice output Rout. It should be noted that the delay time measuring circuit 71, level ratio measuring circuit 72 and configuration comparator 73 are all stopped in operation when a double talk is detected in the control unit 4 (i.e., when the double talk signal DT is output).
B. Operation of the Embodiment
Next, operation of this embodiment will be described.
Firstly, when the far-end talker issues some voice, its content is supplied to the delay time measuring circuit 71 in the form of the far-end voice output Rout. When the near-end voice input Sin, as an echo based on it, is supplied to the delay time measuring circuit 71 before long, the delay time TD is calculated in the delay time measuring circuit 71 and its value is supplied to the delay circuit 75 and the echo path estimation/echo replica generation circuit 3.
As a consequence, the far-end voice output Rout, which is now delayed by TD, is supplied to the level ratio measuring circuit 72. The level ratio measuring circuit 72, in turn, supplies a signal corresponding to a ratio of level between the near-end voice input Sin and the far-end voice output Rout to one end of the multiplier 76. By doing this, the signal Rout', which is now normalized in delay time and level, is supplied to the configuration comparator 73.
In the configuration comparator 73, one of the characteristic data stored in the waveform data base 74, which is most resemble the actual correlation, is selected based on the correlation between the signal Rout' and the near-end voice input Sin. The selected characteristic data is corrected its level by the multiplier 77 and is then supplied to the echo path estimation/echo replica generation circuit 3. By this procedure, an initial value of the tap coefficient of the FIR filter is set. The far-end voice output Rout, which is now delayed by TD, is further supplied to the echo path estimation/echo replica generation circuit 3 through the delay circuit 75.
At the time point when the setting of the delay time TD with respect to the delay circuit 75 and the setting of an initial value of the tap coefficient bi with respect to the FIR filter are all finished in the way as just mentioned, the operation of the initial estimation circuit 7 is finished. Thereafter, in the echo path estimation/echo replica generation circuit 3, a learning identification i s made based on the near-end voice input Sin and the far-end voice output Rout, and the tap coefficient bi is appropriately changed in order to estimate a more correct echo path. In this way, according to the present invention, the delay time of the delay circuit 75 can be set by the delay time measuring circuit 71 and an initial value of the tap coefficient bi in the echo path estimation/echo replica generation circuit 3 can be set by the configuration comparator 73 and the waveform data base 74.
Of course, the estimation of an echo path is not finished by those initial values and a further learning is required in the echo path estimation/echo replica generation circuit 3. However, with the above-mentioned arrangement, the time required for converging the result of learning can be remarkably shortened by rendering the initial values having a somewhat satisfactory degree of correctness to the echo path estimation/echo replica generation circuit 3. Moreover, according to the present invention, since the far-end voice output Rout, which is now delayed by TD, is supplied to the echo path estimation/echo replica generation circuit 3 through the delay circuit 75, load of the learning in the echo path estimation/echo replica generation circuit 3 can be reduced and an echo path estimation can be performed at a higher speed and with a higher accuracy.
As described in the foregoing, according to the echo canceler of this embodiment, since the delay time at the time when an echo replica is generated is pre-set or a comparatively more accurate initial value of each parameter can be set, an echo path estimation can be performed at a high speed and with a high accuracy.
(SIXTH EMBODIMENT)
A. Construction of the Embodiment
FIG. 10 is a block diagram of an important portion of an echo canceler according to one embodiment of the present invention. In this embodiment, a spectrum comparator 81 and a frequency response data base 81 are employed in the place of the configuration comparator 73 and the waveform data base 74 of the fifth embodiment.
The frequency response data base 81 stores data of frequency response characteristics of various kinds of hybrids. Since there are only about dozens kinds of hybrids which are currently used in a telephone line as previously mentioned, it is not a difficult job to store data of the frequency response characteristics of all kinds of hybrids. The spectrum comparator 81 selects one of the frequency response characteristics, which is most resemble the actual characteristic, based on the correlation between the signal Rout' and the near-end voice input Sin. A convertor 83 converts the selected frequency response characteristic to a response characteristic on a time axis. Specifically, the frequency response characteristics are converted to the tap coefficients bi and those tap coefficients bi are set to initial values of the tap coefficients of the FIR filer of the echo path estimation/echo replica generation circuit 3. The spectrum comparator 81 is, as in the case with the configuration comparator 73 of the fifth embodiment, stopped in operation when a double talk is detected (i.e., when a double talk signal DT is output) in the control unit 4.
B. Operation of the Embodiment
Next, operation of this embodiment will be described.
Firstly, when the far-end talker issues some voice, its content is supplied to the delay time measuring circuit 71 in the form of the far-end voice output Rout. When the near-end voice input Sin, as an echo based on it, is supplied to the delay time measuring circuit 71 before long, the delay time TD is calculated in the delay time measuring circuit 71 and its value is supplied to the delay circuit 75 and the echo path estimation/echo replica generation circuit 3.
As a consequence, the far-end voice output Rout, which is now delayed by TD, is supplied to the level ratio measuring circuit 72. The level ratio measuring circuit 72, in turn, supplied a signal corresponding to a ratio of level between the near-end voice input Sin and the far-end voice output Rout to one end of the multiplier 76. By doing this, the signal Rout', which is now normalized in delay time and level, is supplied to the spectrum comparator 81.
In the spectrum comparator 81, one of the characteristic data stored in the frequency response data base 82, which is most resemble the actual characteristic, is selected based on the correlation between the signal Rout' and the near-end voice input Sin. The selected characteristic is converted to the tap coefficient bi through the convertor 83, then corrected its level by the multiplier 77 and then supplied to the echo path estimation/echo replica generation circuit 3. By this procedure, an initial value of the tap coefficient of the FIR filter is set.
At the time point when the setting of the delay time TD with respect to the delay circuit 75 and the setting of an initial value of the tap coefficient bi with respect to the FIR filter are all finished in the way as just mentioned, the operation of the initial estimation circuit 7 is finished. Thereafter, in the echo path estimation/echo replica generation circuit 3, a learning identification is made based on the near-end voice input Sin and the far-end voice output Rout, and the tap coefficient bi is appropriately changed in order to estimate a more correct echo path. In this way, according to this embodiment, the delay time of the delay circuit 75 can be set by the delay time measuring circuit 71 and an initial value of the tap coefficient bi can be set by the delay time measuring circuit 71, frequency response data base 81 and convertor 83.
As described above, according to this embodiment, the time required for converging the result of learning can be remarkably shortened as in the case with the fifth embodiment by rendering the initial values having a somewhat satisfactory degree of correctness to the echo path estimation/echo replica generation circuit 3.
(MODIFIED EMBODIMENT)
It should be noted that the present invention is not limited to the above-mentioned embodiments. For example, many changes and modifications can be made as follows.
1 In any of the above-mentioned embodiments, the present invention is applied to a signal transmission between a mobile telephone and a terrestrial telephone. However, application of the present invention is not limited to this. The present invention can be applied to all communication networks in which a signal transmission is made between a two-wire circuit and a four-wire circuit.
2 In the first and second embodiments, the pseudo noise generator 11 normally generates a pseudo noise. However, an arrangement may also be made, in which a call boundary is received from the switch and no pseudo signal is generated until after the passage of a certain time. Owing to this arrangement, the learning can be finished by the time the near-end talker brings the receiver very close to the ear, so that a pseudo noise can be prevented from being transmitted to the near-end talker at the time a speech is actually made.
3 The pseudo noise generator 11 may generate a pseudo noise when the far-end talker remains in a speechless condition for more than a predetermined time. The reason is that when the far-end talker side is in a speechless condition, transmission of a noise, if any, to the near-end talker side cannot be any hindrance to the speech.
4 The pseudo noise generator 11 may generate a pseudo noise when a cancellation of echo has gone below a predetermined level in the adder 17. The reason is that there is a high possibility of no generation of normal echo replica in the echo replica generator 16 and a re-learning seems to be desirable.
5 The level of a pseudo noise is constant in the first embodiment. However, the level of a pseudo noise may be decreased as the time required for carrying out the correlation operation is increased.
6 In the fourth embodiment, the up- samplers 32, 33 and the down-sampler 34 are constituted of separate elements. In the alternative, they may be constituted of a single element by a digital signal processor (DSP) 35 as shown in FIG. 8. In that case, since it is good enough that the DSP 35 is provided to the conventional construction and a clock to the echo path estimation/echo replica generation circuit 3 is changed to 2 Fs, accuracy of the cancellation characteristic of the echo canceler can be improved by slightly changing the design of the circuit alone.
In other words, since it is good enough that the sampling rate of the input/output signals to the estimation means and the echo replica generator is converted by means of a single signal convertor, a change from the existing devices can be made more easily.
7 In the fourth embodiment, the order of the sampling procedure may be changed in accordance with necessity. Important things here are that the processing timing of the up- samplers 32, 33 and the down-sampler 34 must be time-wise coincident and the output from the down-sampler 34 must be time-wise continuous.
8 The above-mentioned third and fourth embodiments may be combined. FIG. 11 shows one example of a construction of an echo canceler 30 which is comprised of such a combination. As shown in this illustration, a delay circuit 31 is located at a front stage of the up-sampler 32. This delay circuit 31 is adapted to delay the far-end voice input Rin by a transmission delay occurrable between the echo canceler 30 and the hybrid circuit 2. According to this constitutional example, accuracy of the cancellation characteristic of the echo canceler can be more improved by the multiplier effect of the third and fourth embodiments.
9 In the fifth embodiment, the waveform data base 74 stores the tap coefficient bi as the characteristic data. However, the characteristic data is not limited to the tap coefficient bi. For example, an alpha parameter, an LSP parameter, a per-call coefficient or the like, which can indicate the response characteristics of various kinds of hybrids, may be stored as the characteristic data.
10 In the fifth and sixth embodiments, the echo path estimation/echo replica generation circuit 3 is not limited to one which employs the learning identification algorithm. In the alternative, a wide range of algorithms such as Kalman filter and the like may be employed.
11 In the fifth and sixth embodiments, the echo path estimation/echo replica generation circuit 3 may be redesigned such that at first, the level of the echo replica is set to be smaller than the estimated value and it is gradually increased to the original level with the passage of time. The reason is that since it is difficult to estimate an echo replica correctly at the early stage, a high setting of the level of the echo replica signal Y rather results in such an inconvenience as a generation of a noise, etc.
12 In the third to sixth embodiments, an echo path is estimated based on a voice produced by the far-end talker. In the alternative, an echo path may be estimated using other signals. For example, in the case where a calling is issued from the near-end, an echo path may be estimated using a ring back tone, whereas in the case where a calling is issued from the far-end, an echo path may be estimated by transmitting a training signal to the hybrid immediately after the off-hook is made on the near-end side so that an echo path is estimated based thereon. The reason is that owing to the foregoing arrangement, when the near-end talker or the far-end talker actually starts a speech, an accurate echo path estimation is already obtained.
13 In the fifth and sixth embodiments, the far-end voice output Rout, which has been delayed by the delay circuit 75, is supplied to the echo path estimation/echo replica generation circuit 3. In the alternative, a non-delayed far-end voice output Rout may be supplied to the echo path estimation/echo replica generation circuit 3. In that case, the far-end voice output Rout may be delayed by the delay time TD measured in the delay time measuring circuit 71, within the echo path estimation/echo replica generation circuit 3 (specifically, in a numerical FIG. 1, a corresponding tap coefficient bi is brought to "0"). By doing this, even if the delay time TD is varied, the echo replica can be made obedient to the delayed time TD by means of a learning within the echo path estimation/echo replica generation circuit 3.

Claims (14)

We claim:
1. An echo canceler employed in a communication line network including a first transmission line for transmitting a voice of a four-wire side and a second transmission line for transmitting a voice of a second-wire side, said echo canceler comprising:
training signal generator means for generating a training signal and supplying the same to said first transmission line;
coefficient calculator means for calculating a coefficient necessary for generating an echo replica based on a correlation established between the training signal supplied to said first transmission line and a signal of said second transmission line;
storage means for storing response characteristics of a plurality of hybrids;
selector means for selecting one of said response characteristics based on a relation between a transmission signal propagated from said four-wire side to said two-wire side and an echo signal propagated from said two-wire side to said four-wire side;
echo replica generator means for estimating an echo path by a predetermined learning algorithm and generating an echo replica by appropriately delaying the transmission signal and delaying the signal in accordance with a predetermined parameter; and
initial parameter setting means for setting an initial value of said parameter based on the response characteristic selected by said selector means.
2. An echo canceler according to claim 1, further comprising:
delay time measuring means for measuring a delay time of the echo signal with respect to the transmission signal based on the transmission signal transmitted from said four-wire side to said two-wire side and the echo signal transmitted from said four-wire side to said two-wire side; and
delay setting means for setting the delay time in said echo replica generator means to a delay time measured by said delay time measuring means.
3. An echo canceler employed in a communication line network for performing a transmission between a voice transmitted through a four-wire side and a voice transmitted through a two-wire side, said echo canceler comprising:
storage means for storing response characteristics of a plurality of hybrids;
selector means for selecting one of said response characteristics based on a relation between a transmission signal propagated from said four-wire side to said two-wire side and an echo signal propagated from said two-wire side to said four-wire side;
echo replica generator means for estimating an echo path by a predetermined learning algorithm and generating an echo replica by appropriately delaying the transmission signal and transfiguring the signal in accordance with a predetermined parameter; and
initial parameter setting means for setting an initial value of said parameter based on the response characteristic selected by said selector means.
4. An echo path estimating method applicable to an echo canceler employed in a communication line network including a first transmission line for transmitting a voice of a four-wire side and a second transmission line for transmitting a voice of a second-wire side, said echo pass estimating method comprising:
a training signal output step for generating a training signal and supplying the same to said first transmission line;
a coefficient calculation step for calculating a coefficient necessary for generating an echo replica based on a correlation established between the training signal supplied to said first transmission line and a signal of said second transmission line;
storing response characteristics of a plurality of hybrids;
selecting one of said response characteristics based on a relation between a transmission signal propagated from said four-wire side to said two-wire side and an echo signal propagated from said two-wire side to said four-wire side;
estimating an echo path by a predetermined learning algorithm and generating an echo replica by appropriately delaying the transmission signal and transfiguring the signal in accordance with a predetermined parameter; and
setting an initial value of said parameter based on the response characteristic selected by said selector means.
5. An echo path estimating method for estimating an echo path with respect to a communication line network for performing a transmission between a voice transmitted through a four-wire side and a voice transmitted through a two-wire side, said method comprising the steps of:
storing response characteristics of a plurality of hybrids;
selecting one of said response characteristics based on a relation between a transmission signal propagated from said four-wire side to said two-wire side and an echo signal propagated from said two-wire side to said four-wire side;
estimating an echo path by a predetermined learning algorithm and generating an echo replica by appropriately delaying the transmission signal and transfiguring the signal in accordance with a predetermined parameter; and
setting an initial value of said parameter based on the response characteristic selected by said selector means.
6. An echo canceler employed in a communication line network including a first transmission line for transmitting a voice of a four-wire side and a second transmission line for transmitting a voice of a second-wire side, said echo canceler comprising:
training signal generator means for generating a training signal and supplying the same to said first transmission line;
coefficient calculator means for calculating a coefficient necessary for generating an echo replica based on a correlation established between the training signal supplied to said first transmission line and a signal of said second transmission line;
storage means for storing frequency characteristics of a plurality of hybrids;
selector means for selecting one of said frequency response characteristics based on a relation between a transmission signal propagated from said four-wire side to said two-wire side and an echo signal propagated from said two-wire side to said four-wire side;
echo replica generator means for estimating an echo path by a predetermined learning algorithm and generating an echo replica by appropriately delaying the transmission signal and transfiguring the signal in accordance with a predetermined parameter; and
initial parameter set means for setting an initial value of said parameter based on the frequency response characteristic selected by said selector means.
7. An echo canceler according to claim 6, further comprising:
delay time measuring means for measuring a delay time of the echo signal with respect to the transmission signal based on the transmission signal transmitted from said four-wire side to said two-wire side and the echo signal transmitted from said two-wire side to said four-wire side; and
delay setting means for setting a delay time in said echo replica generator means to a delay time measured by said delay time measuring means.
8. An echo canceler employed in a communication line network for performing a transmission between a voice transmitted through a four-wire side and a voice transmitted through a two-wire side, said echo canceler comprising:
storage means for storing frequency characteristics of a plurality of hybrids;
selector means for selecting one of said frequency response characteristics based on a relation between a transmission signal propagated from said four-wire side to said two-wire side and an echo signal propagated from said two-wire side to said four-wire side;
echo replica generator means for estimating an echo path by a predetermined learning algorithm and generating an echo replica by appropriately delaying the transmission signal and transfiguring the signal in accordance with a predetermined parameter; and
initial parameter set means for setting an initial value of said parameter based on the frequency response characteristic selected by said selector means.
9. An echo path estimating method applicable to an echo canceler employed in a communication line network including a first transmission line for transmitting a voice of a four-wire side and a second transmission line for transmitting a voice of a second-wire side, said echo pass estimating method comprising:
a training signal output step for generating a training signal and supplying the same to said first transmission line;
a coefficient calculation step for calculating a coefficient necessary for generating an echo replica based on a correlation established between the training signal supplied to said first transmission line and a signal of said second transmission line;
storing frequency characteristics of a plurality of hybrids;
selecting one of said frequency response characteristics based on a relation between a transmission signal propagated from said four-wire side to said two-wire side and an echo signal propagated from said two-wire side to said four-wire side;
estimating an echo path by a predetermined learning algorithm and generating an echo replica by appropriately delaying the transmission signal and transfiguring the signal in accordance with a predetermined parameter; and
setting an initial value of said parameter based on the frequency response characteristic selected by said selector means.
10. An echo path estimating method for estimating an echo path with respect to a communication line network for performing a transmission between a voice transmitted through a four-wire side and a voice transmitted through a two-wire side, said method comprising the steps of:
storing frequency response characteristics of a plurality of hybrids;
selecting one of said frequency response characteristics based on a relation between a transmission signal propagated from said four-wire side to said two-wire side and an echo signal propagated from said two-wire side to said four-wire side;
estimating an echo path by a predetermined learning algorithm and generating an echo replica by appropriately delaying the transmission signal and transfiguring the signal in accordance with a predetermined parameter; and
setting an initial value of said parameter based on the frequency response characteristic selected by said selector means.
11. An echo canceler employed in a communication line network including a first transmission line for transmitting a voice of a four-wire side and a second transmission line for transmitting a voice of a second-wire side, said echo canceler comprising:
training signal generator means for generating a training signal and supplying the same to said first transmission line;
coefficient calculator means for calculating a coefficient necessary for generating an echo replica based on a correlation established between the training signal supplied to said first transmission line and a signal of said second transmission line;
first sampling means for up-sampling a first digital voice signal corresponding to the voice on said four-wire side;
second sampling means for up-sampling a second digital voice signal corresponding to the voice on said two-wire side at a same sampling rate as said first sampling means;
estimation means for estimating, on the basis of sampling results of the first and second up-sampler means, a response characteristic of a hybrid circuit for converting the four-wire side and the two-wire side;
echo replica generator means for artificially generating an echo caused by said hybrid circuit by means of calculation using the estimated result and the sampled result by said sampling means;
third sampling means for down-sampling an output signal of said echo replica generator means; and
substraction means for subtracting an output signal of said third sampling means from said second digital voice signal.
12. An echo canceler according to claim 11, further comprising delay means located at a front stage of said estimation means and adapted to delay a predetermined time.
13. An echo canceler according to claim 11 or 12, wherein the up-sampling or down-sampling by said first to third sampling means is performed by a single signal convertor means, said signal convertor means converting a sampling rate of the input/output signal to said estimation means and said echo replica generator means.
14. An echo canceler employed in a communication line network for performing a transmission between a voice transmitted through a four-wire side and a voice transmitted through a two-wire side, said echo canceler comprising:
first sampling means for up-sampling a first digital voice signal corresponding to the voice on said four-wire side;
second sampling means for up-sampling a second digital voice signal corresponding to the voice on said two-wire side at a same sampling rate as said first sampling means;
estimation means for estimating, on the basis of sampling results of the first and second up-sampler means, a response characteristic of a hybrid circuit for converting the four-wire side and the two-wire side;
echo replica generator means for artificially generating an echo caused by said hybrid circuit by means of calculation using the estimated result and the sampled result by said sampling means;
third sampling means for down-sampling an output signal of said echo replica generator means; and
substraction means for subtracting an output signal of said third sampling means from said second digital voice signal.
US08/507,507 1994-05-06 1995-05-02 Echo canceler and echo path estimating method Expired - Fee Related US5859907A (en)

Applications Claiming Priority (8)

Application Number Priority Date Filing Date Title
JP11764094A JPH07303067A (en) 1994-05-06 1994-05-06 Echo canceler
JP94117640 1994-05-06
JP11759294A JPH07303064A (en) 1994-05-07 1994-05-07 Echo canceler
JP11758994A JPH07303061A (en) 1994-05-07 1994-05-07 Echo canceler and echo path estimating method
JP6117590A JPH07303062A (en) 1994-05-07 1994-05-07 Echo canceler and echo path estimating method
JP11759394A JP3212796B2 (en) 1994-05-07 1994-05-07 Echo canceller
JP94117589 1994-05-07
PCT/JP1995/000866 WO1995031052A1 (en) 1994-05-06 1995-05-02 Echo canceler and echo path estimating method

Publications (1)

Publication Number Publication Date
US5859907A true US5859907A (en) 1999-01-12

Family

ID=27526770

Family Applications (1)

Application Number Title Priority Date Filing Date
US08/507,507 Expired - Fee Related US5859907A (en) 1994-05-06 1995-05-02 Echo canceler and echo path estimating method

Country Status (8)

Country Link
US (1) US5859907A (en)
EP (3) EP1184998B1 (en)
KR (1) KR100233077B1 (en)
CN (3) CN1232048C (en)
AT (1) ATE235763T1 (en)
CA (1) CA2162570C (en)
DE (3) DE69530040T2 (en)
WO (1) WO1995031052A1 (en)

Cited By (27)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6061444A (en) * 1994-05-07 2000-05-09 Ntt Mobile Communications Network, Inc. Echo canceler and method for learning for the same
US6081593A (en) * 1997-02-25 2000-06-27 Samsung Electronics Co., Ltd. Method and apparatus for transmitting white noise for echo cancellation in a voice mail system
US6141406A (en) * 1997-03-27 2000-10-31 T-Netix, Inc. Method and apparatus for detecting a secondary destination of a telephone call based on changes in the telephone signal path
WO2000070854A1 (en) * 1999-05-18 2000-11-23 Telefonaktiebolaget Lm Ericsson (Publ) Methods and apparatus for improving adaptive filter initialization and recovery in bi-directional communications systems
US6181753B1 (en) * 1997-04-30 2001-01-30 Oki Electric Industry Co., Ltd. Echo/noise canceler with delay compensation
US6272106B1 (en) * 1994-05-06 2001-08-07 Nit Mobile Communications Network, Inc. Method and device for detecting double-talk, and echo canceler
US6324170B1 (en) * 1998-09-10 2001-11-27 Nortel Networks Limited Echo controller with compensation for variable delay networks
US20020071400A1 (en) * 2000-12-08 2002-06-13 Claymens Lee Apparatus for a digital echo canceller and method therefor
US20020172350A1 (en) * 2001-05-15 2002-11-21 Edwards Brent W. Method for generating a final signal from a near-end signal and a far-end signal
US6516050B1 (en) * 1999-02-25 2003-02-04 Mitsubishi Denki Kabushiki Kaisha Double-talk detecting apparatus, echo canceller using the double-talk detecting apparatus and echo suppressor using the double-talk detecting apparatus
US6549627B1 (en) * 1998-01-30 2003-04-15 Telefonaktiebolaget Lm Ericsson Generating calibration signals for an adaptive beamformer
US6618479B1 (en) * 1997-09-04 2003-09-09 Matsushita Electric Industrial Co., Ltd. Echo canceler
WO2003092184A1 (en) * 2002-04-26 2003-11-06 Global Ip Sound Inc Echo cancellation
US6707912B2 (en) * 1999-03-11 2004-03-16 Motorola, Inc. Method and apparatus for setting a step size for an adaptive filter coefficient of an echo canceller
US6763099B1 (en) * 1999-11-05 2004-07-13 T-Netix, Inc. Advanced three way call detection system and method using spread spectrum techniques
US20050014491A1 (en) * 1997-03-27 2005-01-20 T-Netix, Inc. Method and apparatus for detecting a secondary destination of a telephone call based on changes in the telephone signal path
US7062039B1 (en) * 1999-05-27 2006-06-13 Telefonaktiebolaget Lm Ericsson Methods and apparatus for improving adaptive filter performance by inclusion of inaudible information
US20070242658A1 (en) * 2006-04-13 2007-10-18 Evercom Systems, Inc. Unauthorized call activity detection and prevention systems and methods for a voice over internet protocol environment
US20100074433A1 (en) * 2008-09-22 2010-03-25 Microsoft Corporation Multichannel Acoustic Echo Cancellation
US7899167B1 (en) 2003-08-15 2011-03-01 Securus Technologies, Inc. Centralized call processing
US8000269B1 (en) 2001-07-13 2011-08-16 Securus Technologies, Inc. Call processing with voice over internet protocol transmission
US8340260B1 (en) 2003-08-15 2012-12-25 Securus Technologies, Inc. Inmate management and call processing systems and methods
US9560193B1 (en) 2002-04-29 2017-01-31 Securus Technologies, Inc. Systems and methods for detecting a call anomaly using biometric identification
US20170244839A1 (en) * 2000-07-12 2017-08-24 Michael Pascazi System and method for cell phone to cell phone signal transmission via the internet
US9990683B2 (en) 2002-04-29 2018-06-05 Securus Technologies, Inc. Systems and methods for acquiring, accessing, and analyzing investigative information
US10115080B2 (en) 2002-04-29 2018-10-30 Securus Technologies, Inc. System and method for proactively establishing a third-party payment account for services rendered to a resident of a controlled-environment facility
US10796392B1 (en) 2007-05-22 2020-10-06 Securus Technologies, Llc Systems and methods for facilitating booking, bonding and release

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6317419B1 (en) 1998-06-10 2001-11-13 Conexant Systems, Inc. Method and apparatus for training an echo canceler in a PCM modem context
KR100399902B1 (en) * 2001-05-02 2003-09-29 주식회사 하이닉스반도체 Active hybrid circuit for high speed wired communication
KR20030091362A (en) * 2002-05-27 2003-12-03 엘지전자 주식회사 Line echo canceller
KR100502414B1 (en) * 2002-11-22 2005-07-19 삼성전자주식회사 Echo canceller of adsl system and method for training thereof
JP4613705B2 (en) * 2005-06-15 2011-01-19 沖電気工業株式会社 Echo canceller
JP5092580B2 (en) * 2007-06-26 2012-12-05 ソニー株式会社 Digital signal processing apparatus, digital signal processing method, and digital signal processing program
CN101350798B (en) * 2008-08-29 2010-06-16 北京创毅视讯科技有限公司 Apparatus and method for eliminating echo, transmit-receive equipment containing the apparatus
JP6155882B2 (en) * 2013-06-17 2017-07-05 沖電気工業株式会社 Echo suppression device and program
CN104270538B (en) * 2014-08-22 2016-05-25 北京电子科技学院 Dual transformer echo cancellation digital telephone cipher machine
KR102076760B1 (en) * 2018-09-19 2020-02-12 한양대학교 산학협력단 Method for cancellating nonlinear acoustic echo based on kalman filtering using microphone array

Citations (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS5890832A (en) * 1981-11-26 1983-05-30 Toshiba Corp Echo canceller
US4562312A (en) * 1983-02-17 1985-12-31 At&T Bell Laboratories Subsampling delay estimator for an echo canceler
US4571720A (en) * 1983-02-25 1986-02-18 Telecommunications Radioelectriques Et Telephoniques T.R.T. Method and apparatus for reducing the convergence time of an echo canceller
US4582963A (en) * 1982-07-29 1986-04-15 Rockwell International Corporation Echo cancelling using adaptive bulk delay and filter
US4621173A (en) * 1982-06-14 1986-11-04 Telecommunications Radioelectriques et Telephoniques T.R.I Method of reducing the convergence time of an echo canceller and apparatus for carrying out said method
JPS62116025A (en) * 1985-11-15 1987-05-27 Nippon Telegr & Teleph Corp <Ntt> Echo canceler
EP0391715A2 (en) * 1989-04-05 1990-10-10 Codex Corporation Echo canceller
EP0467163A1 (en) * 1990-07-06 1992-01-22 Fujitsu Limited Hybrid ciruit having a two-wire/four-wire converting function
JPH0465922A (en) * 1990-07-02 1992-03-02 Fujitsu Ltd Echo canceller
US5131032A (en) * 1989-03-13 1992-07-14 Hitachi, Ltd. Echo canceller and communication apparatus employing the same
EP0498369A2 (en) * 1991-02-04 1992-08-12 Telebit Corporation Method and apparatus for echo cancellation of a multicarrier signal
US5151937A (en) * 1989-07-12 1992-09-29 Fujitsu Limited Adaptive echo canceller
US5305309A (en) * 1989-12-06 1994-04-19 Fujitsu Limited Echo canceller
US5327496A (en) * 1993-06-30 1994-07-05 Iowa State University Research Foundation, Inc. Communication device, apparatus, and method utilizing pseudonoise signal for acoustical echo cancellation
US5604799A (en) * 1994-11-11 1997-02-18 Nec Corporation Switching type echo canceler
US5696819A (en) * 1993-01-29 1997-12-09 Kabushiki Kaisha Toshiba Speech communication apparatus

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
AU4695485A (en) * 1984-09-21 1986-03-27 Alcatel N.V. Digital hybrid
JPS61184934A (en) * 1985-02-12 1986-08-18 Nippon Telegr & Teleph Corp <Ntt> Digital subscriber line two-way transmitter
JPS6232727A (en) * 1985-08-05 1987-02-12 Fujitsu Ltd Echo canceller device
JPH0638589B2 (en) * 1985-10-25 1994-05-18 日本電信電話株式会社 Eco-Cancer
DE3585034D1 (en) * 1985-10-30 1992-02-06 Ibm METHOD FOR DETERMINING A FLAT Echo Path Delay, and Echo Compensator Using This Method.
EP0403716B1 (en) * 1989-06-22 1995-09-13 International Business Machines Corporation Echo cancelling device providing adjustment of the echo canceller coefficients during full-duplex transmission

Patent Citations (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS5890832A (en) * 1981-11-26 1983-05-30 Toshiba Corp Echo canceller
US4621173A (en) * 1982-06-14 1986-11-04 Telecommunications Radioelectriques et Telephoniques T.R.I Method of reducing the convergence time of an echo canceller and apparatus for carrying out said method
US4582963A (en) * 1982-07-29 1986-04-15 Rockwell International Corporation Echo cancelling using adaptive bulk delay and filter
US4562312A (en) * 1983-02-17 1985-12-31 At&T Bell Laboratories Subsampling delay estimator for an echo canceler
US4571720A (en) * 1983-02-25 1986-02-18 Telecommunications Radioelectriques Et Telephoniques T.R.T. Method and apparatus for reducing the convergence time of an echo canceller
JPS62116025A (en) * 1985-11-15 1987-05-27 Nippon Telegr & Teleph Corp <Ntt> Echo canceler
US5131032A (en) * 1989-03-13 1992-07-14 Hitachi, Ltd. Echo canceller and communication apparatus employing the same
EP0391715A2 (en) * 1989-04-05 1990-10-10 Codex Corporation Echo canceller
US5151937A (en) * 1989-07-12 1992-09-29 Fujitsu Limited Adaptive echo canceller
US5305309A (en) * 1989-12-06 1994-04-19 Fujitsu Limited Echo canceller
JPH0465922A (en) * 1990-07-02 1992-03-02 Fujitsu Ltd Echo canceller
EP0467163A1 (en) * 1990-07-06 1992-01-22 Fujitsu Limited Hybrid ciruit having a two-wire/four-wire converting function
EP0498369A2 (en) * 1991-02-04 1992-08-12 Telebit Corporation Method and apparatus for echo cancellation of a multicarrier signal
US5696819A (en) * 1993-01-29 1997-12-09 Kabushiki Kaisha Toshiba Speech communication apparatus
US5327496A (en) * 1993-06-30 1994-07-05 Iowa State University Research Foundation, Inc. Communication device, apparatus, and method utilizing pseudonoise signal for acoustical echo cancellation
US5604799A (en) * 1994-11-11 1997-02-18 Nec Corporation Switching type echo canceler

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
Patent Abstracts of Japan, vol. 11, No. 11, E 470. *
Patent Abstracts of Japan, vol. 11, No. 11, E-470.
Patent Abstracts of Japan, vol. 11, No. 212, E 522. *
Patent Abstracts of Japan, vol. 11, No. 212, E-522.
Patent Abstracts of Japan, vol. 11, No. 306, E 546. *
Patent Abstracts of Japan, vol. 11, No. 306, E-546.

Cited By (37)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6272106B1 (en) * 1994-05-06 2001-08-07 Nit Mobile Communications Network, Inc. Method and device for detecting double-talk, and echo canceler
US6061444A (en) * 1994-05-07 2000-05-09 Ntt Mobile Communications Network, Inc. Echo canceler and method for learning for the same
US6081593A (en) * 1997-02-25 2000-06-27 Samsung Electronics Co., Ltd. Method and apparatus for transmitting white noise for echo cancellation in a voice mail system
US6141406A (en) * 1997-03-27 2000-10-31 T-Netix, Inc. Method and apparatus for detecting a secondary destination of a telephone call based on changes in the telephone signal path
US7136471B2 (en) * 1997-03-27 2006-11-14 T-Netix, Inc. Method and apparatus for detecting a secondary destination of a telephone call based on changes in the telephone signal path
US20050014491A1 (en) * 1997-03-27 2005-01-20 T-Netix, Inc. Method and apparatus for detecting a secondary destination of a telephone call based on changes in the telephone signal path
US6181753B1 (en) * 1997-04-30 2001-01-30 Oki Electric Industry Co., Ltd. Echo/noise canceler with delay compensation
US6618479B1 (en) * 1997-09-04 2003-09-09 Matsushita Electric Industrial Co., Ltd. Echo canceler
US6549627B1 (en) * 1998-01-30 2003-04-15 Telefonaktiebolaget Lm Ericsson Generating calibration signals for an adaptive beamformer
US6324170B1 (en) * 1998-09-10 2001-11-27 Nortel Networks Limited Echo controller with compensation for variable delay networks
US6516050B1 (en) * 1999-02-25 2003-02-04 Mitsubishi Denki Kabushiki Kaisha Double-talk detecting apparatus, echo canceller using the double-talk detecting apparatus and echo suppressor using the double-talk detecting apparatus
US6707912B2 (en) * 1999-03-11 2004-03-16 Motorola, Inc. Method and apparatus for setting a step size for an adaptive filter coefficient of an echo canceller
WO2000070854A1 (en) * 1999-05-18 2000-11-23 Telefonaktiebolaget Lm Ericsson (Publ) Methods and apparatus for improving adaptive filter initialization and recovery in bi-directional communications systems
US6560332B1 (en) 1999-05-18 2003-05-06 Telefonaktiebolaget Lm Ericsson (Publ) Methods and apparatus for improving echo suppression in bi-directional communications systems
US7062039B1 (en) * 1999-05-27 2006-06-13 Telefonaktiebolaget Lm Ericsson Methods and apparatus for improving adaptive filter performance by inclusion of inaudible information
US6763099B1 (en) * 1999-11-05 2004-07-13 T-Netix, Inc. Advanced three way call detection system and method using spread spectrum techniques
US10708442B2 (en) * 2000-07-12 2020-07-07 Michael Pascazi System and method for cell phone to cell phone signal transmission via the internet
US20170244839A1 (en) * 2000-07-12 2017-08-24 Michael Pascazi System and method for cell phone to cell phone signal transmission via the internet
US6804204B2 (en) * 2000-12-08 2004-10-12 Topic Semiconductor Corp. Apparatus for a digital echo canceller and method therefor
US20020071400A1 (en) * 2000-12-08 2002-06-13 Claymens Lee Apparatus for a digital echo canceller and method therefor
US20020172350A1 (en) * 2001-05-15 2002-11-21 Edwards Brent W. Method for generating a final signal from a near-end signal and a far-end signal
US8000269B1 (en) 2001-07-13 2011-08-16 Securus Technologies, Inc. Call processing with voice over internet protocol transmission
US20050220043A1 (en) * 2002-04-26 2005-10-06 Global Ip Sound Inc Echo cancellation
US7366118B2 (en) 2002-04-26 2008-04-29 Global Ip Solutions Inc. Echo cancellation
WO2003092184A1 (en) * 2002-04-26 2003-11-06 Global Ip Sound Inc Echo cancellation
US9560193B1 (en) 2002-04-29 2017-01-31 Securus Technologies, Inc. Systems and methods for detecting a call anomaly using biometric identification
US9990683B2 (en) 2002-04-29 2018-06-05 Securus Technologies, Inc. Systems and methods for acquiring, accessing, and analyzing investigative information
US10115080B2 (en) 2002-04-29 2018-10-30 Securus Technologies, Inc. System and method for proactively establishing a third-party payment account for services rendered to a resident of a controlled-environment facility
US10178224B2 (en) 2002-04-29 2019-01-08 Securus Technologies, Inc. Systems and methods for detecting a call anomaly using biometric identification
US8340260B1 (en) 2003-08-15 2012-12-25 Securus Technologies, Inc. Inmate management and call processing systems and methods
US7899167B1 (en) 2003-08-15 2011-03-01 Securus Technologies, Inc. Centralized call processing
US10740861B1 (en) 2003-11-24 2020-08-11 Securus Technologies, Inc. Systems and methods for acquiring, accessing, and analyzing investigative information
US7916845B2 (en) 2006-04-13 2011-03-29 Securus Technologies, Inc. Unauthorized call activity detection and prevention systems and methods for a Voice over Internet Protocol environment
US20070242658A1 (en) * 2006-04-13 2007-10-18 Evercom Systems, Inc. Unauthorized call activity detection and prevention systems and methods for a voice over internet protocol environment
US10796392B1 (en) 2007-05-22 2020-10-06 Securus Technologies, Llc Systems and methods for facilitating booking, bonding and release
US8605890B2 (en) * 2008-09-22 2013-12-10 Microsoft Corporation Multichannel acoustic echo cancellation
US20100074433A1 (en) * 2008-09-22 2010-03-25 Microsoft Corporation Multichannel Acoustic Echo Cancellation

Also Published As

Publication number Publication date
CA2162570C (en) 2001-04-10
DE69533375D1 (en) 2004-09-16
DE69535116T2 (en) 2007-02-15
CN1185800C (en) 2005-01-19
EP1202469A2 (en) 2002-05-02
CN1404232A (en) 2003-03-19
CN1084976C (en) 2002-05-15
EP1184998A2 (en) 2002-03-06
CN1128591A (en) 1996-08-07
CN1396718A (en) 2003-02-12
DE69530040T2 (en) 2004-02-12
ATE235763T1 (en) 2003-04-15
EP1202469B1 (en) 2004-08-11
KR960702965A (en) 1996-05-23
EP1184998A9 (en) 2003-12-17
EP1202469A3 (en) 2003-04-02
KR100233077B1 (en) 1999-12-01
EP1184998B1 (en) 2006-07-19
DE69530040D1 (en) 2003-04-30
EP0708537B1 (en) 2003-03-26
WO1995031052A1 (en) 1995-11-16
EP0708537A1 (en) 1996-04-24
DE69533375T2 (en) 2005-07-28
EP0708537A4 (en) 1997-08-06
CN1232048C (en) 2005-12-14
CA2162570A1 (en) 1995-11-16
EP1184998A3 (en) 2003-08-13
DE69535116D1 (en) 2006-08-31

Similar Documents

Publication Publication Date Title
US5859907A (en) Echo canceler and echo path estimating method
AU680981B2 (en) Method for determining the location of echo in an echo cancellar
US5263020A (en) Echo canceller
EP0868787B1 (en) Method and device for echo cancellation using power estimation in a residual signal
EP0667700B1 (en) Echo cancelling method and apparatus using fast projection scheme
US5315585A (en) Echo canceller using two residual echoes
US5852661A (en) Adaptive echo cancellation used with echo suppression to reduce short and long duration echoes
US5896452A (en) Multi-channel echo canceler and method using convolution of two training signals
US5247512A (en) Echo canceller
CA2162571C (en) Echo canceler and method for learning for the same
JPH0648790B2 (en) Controller for echo canceller and center clipper
US5951626A (en) Adaptive filter
EP0855110B1 (en) An adaptive echo cancellation method
US7386120B2 (en) Echo canceller having a frequency domain active region location estimator
CA2514434C (en) Method of adaptation step control in echo cancellers
JPH07303067A (en) Echo canceler
JP3094634B2 (en) Echo removal method and echo removal device
JPH07303061A (en) Echo canceler and echo path estimating method
JPH08213939A (en) Echo canceler
JPH07303062A (en) Echo canceler and echo path estimating method
JPH07303063A (en) Echo canceler and echo path estimating method

Legal Events

Date Code Title Description
AS Assignment

Owner name: NTT MOBILE COMMUNICATIONS NETWORK, JAPAN

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:KAWAHARA, TOSHIRO;MIKI, TOSHIO;SUDA, HIROHITO;AND OTHERS;REEL/FRAME:007886/0140

Effective date: 19950801

CC Certificate of correction
FEPP Fee payment procedure

Free format text: PAYOR NUMBER ASSIGNED (ORIGINAL EVENT CODE: ASPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

FPAY Fee payment

Year of fee payment: 4

FEPP Fee payment procedure

Free format text: PAYER NUMBER DE-ASSIGNED (ORIGINAL EVENT CODE: RMPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

Free format text: PAYOR NUMBER ASSIGNED (ORIGINAL EVENT CODE: ASPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

FPAY Fee payment

Year of fee payment: 8

REMI Maintenance fee reminder mailed
LAPS Lapse for failure to pay maintenance fees
STCH Information on status: patent discontinuation

Free format text: PATENT EXPIRED DUE TO NONPAYMENT OF MAINTENANCE FEES UNDER 37 CFR 1.362

FP Lapsed due to failure to pay maintenance fee

Effective date: 20110112